Merge pull request #141 from Hashara/hashara

Hashara

Author: Hashara

.github/workflows/ci.yaml

@@ -178,7 +178,7 @@ jobs:
 
   build-macos-x86_64:
     name: Mac OS x86-64
-    runs-on: macos-13 # Mac OS 14 Does Not Support x86-64
+    runs-on: macos-15-intel # replaces deprecated macos-13
 
     steps:
       - name: Checkout
@@ -196,7 +196,7 @@ jobs:
           cd build
           export CPPFLAGS="-I/usr/local/opt/libomp/include"
           export CXXFLAGS="-I/usr/local/opt/libomp/include"
-          cmake .. -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++
+          cmake .. -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ -DEIGEN3_INCLUDE_DIR=$(brew --prefix eigen)/include/eigen3
           make -j
           make package
           file iqtree3 | grep x86_64
@@ -249,7 +249,7 @@ jobs:
           cd build
           export CPPFLAGS="-I/opt/homebrew/opt/libomp/include"
           export CXXFLAGS="-I/opt/homebrew/opt/libomp/include"
-          cmake .. -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++
+          cmake .. -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ -DEIGEN3_INCLUDE_DIR=$(brew --prefix eigen)/include/eigen3
           make -j
           make package
           file iqtree3 | grep arm64

.gitignore

@@ -34,3 +34,4 @@ vectorclass/
 zlib-1.2.7/
 /.direnv/
 /.envrc
+softwipe_build

CMakeLists.txt

@@ -120,10 +120,13 @@ find_package(Backtrace)
 
 # The version number.
 set (iqtree_VERSION_MAJOR 3)
-set (iqtree_VERSION_MINOR 0)
-set (iqtree_VERSION_PATCH ".1")
+set (iqtree_VERSION_MINOR 1)
+set (iqtree_VERSION_PATCH ".0")
 
 option(BUILD_SHARED_LIBS "Build Shared Libraries" OFF)
+option(USE_CUDA "Enable CUDA support" OFF)
+option(USE_OPENACC "Enable OpenACC support" OFF)
+option(USE_OPENACC_PROFILE "Enable profiling instrumentation in OpenACC kernels" OFF)
 
 if (CMAKE_C_COMPILER MATCHES "mpic")
     set(IQTREE_FLAGS "${IQTREE_FLAGS} mpi")
@@ -227,6 +230,13 @@ if (APPLE)
         set (__ARM_NEON "TRUE")
         set (NEON 1)
     endif()
+    execute_process(
+      COMMAND sw_vers -productVersion
+      OUTPUT_VARIABLE MACOS_VERSION
+      OUTPUT_STRIP_TRAILING_WHITESPACE
+    )
+    string(REGEX MATCH "^([0-9]+)" MACOS_MAJOR_VERSION "${MACOS_VERSION}")
+    message("macOS version: ${MACOS_VERSION} (major: ${MACOS_MAJOR_VERSION})")
 elseif (UNIX AND NOT APPLE)  # Unix and Linux
     execute_process(
       COMMAND uname -m
@@ -449,10 +459,86 @@ if (MSVC)
 endif()
 
 if (NVHPC)
+
     set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -noswitcherror")
-    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -cuda") # to disable AVX512 intrinsics
+#    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -cuda") # to disable AVX512 intrinsics
     set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Minfo=accel")
 endif()
+if (USE_CUDA)
+    message("CUDA : Yes")
+    enable_language(CUDA)
+    add_definitions(-DUSE_CUDA)
+    set(CMAKE_CUDA_STANDARD 14)
+
+    # Set CUDA architectures if not already specified
+    if (NOT DEFINED CMAKE_CUDA_ARCHITECTURES)
+        set(CMAKE_CUDA_ARCHITECTURES "70;75;80;86;89;90")
+    endif()
+
+    # Manual CUDA runtime detection — avoids FindCUDAToolkit's Threads dependency
+    # enable_language(CUDA) already found nvcc and set CMAKE_CUDA_* variables.
+    find_library(CUDART_LIBRARY cudart
+        HINTS ${CMAKE_CUDA_IMPLICIT_LINK_DIRECTORIES}
+              ENV CUDA_HOME
+              ENV CUDA_PATH
+              ENV CUDAToolkit_ROOT
+        PATH_SUFFIXES lib64 lib lib/x86_64-linux-gnu)
+    if (CUDART_LIBRARY)
+        message(STATUS "Found cudart: ${CUDART_LIBRARY}")
+    else()
+        message(FATAL_ERROR "Could not find cudart library. Set CUDA_HOME or pass -DCMAKE_CUDA_IMPLICIT_LINK_DIRECTORIES=<cuda>/lib64")
+    endif()
+
+    if (NVHPC)
+        set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -DUSE_CUDA")
+    endif()
+
+    if (CMAKE_CUDA_TOOLKIT_INCLUDE_DIRECTORIES)
+        include_directories(${CMAKE_CUDA_TOOLKIT_INCLUDE_DIRECTORIES})
+    endif()
+    if (CMAKE_CUDA_IMPLICIT_LINK_DIRECTORIES)
+        link_directories(${CMAKE_CUDA_IMPLICIT_LINK_DIRECTORIES})
+    endif()
+else ()
+    message("CUDA : No")
+endif()
+
+if (USE_OPENACC STREQUAL "ON")
+    message("OpenACC       : Yes")
+    add_definitions(-DUSE_OPENACC)
+    if (GCC)
+        set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fopenacc -O3")
+    elseif (NVHPC)
+        set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -acc -O3 -fast -Minfo=accel")
+        # Target specific GPU architecture for optimal code generation.
+        # -gpu=ccXX generates arch-specific instructions instead of generic PTX.
+        # Examples: cc70 (V100), cc80 (A100), cc89 (RTX 4090), cc90 (H100)
+        # Multiple targets: -DGPU_ARCH="cc70,cc80,cc90"
+        if (DEFINED GPU_ARCH)
+            set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -gpu=${GPU_ARCH},maxregcount:128")
+            message("OpenACC GPU   : ${GPU_ARCH}")
+        else()
+            set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -gpu=cc70,cc80,cc90,maxregcount:128")
+            message("OpenACC GPU   : cc70,cc80,cc90 (default)")
+        endif()
+    endif()
+    else()
+    message("OpenACC       : NONE")
+endif()
+
+if (USE_OPENACC_PROFILE STREQUAL "ON")
+    if (NOT USE_OPENACC STREQUAL "ON")
+        message(FATAL_ERROR "USE_OPENACC_PROFILE requires USE_OPENACC=ON")
+    endif()
+    add_definitions(-DUSE_OPENACC_PROFILE)
+    # Replace -O3 -fast with -O2 -g for debuggable profiling builds
+    string(REPLACE "-O3" "-O2" CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}")
+    string(REPLACE "-fast" "" CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}")
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -g")
+    message("OpenACC Prof  : Yes (-O2 -g, debug symbols enabled)")
+else()
+    message("OpenACC Prof  : No")
+endif()
 
 # enable link time optimization
 if (IQTREE_FLAGS MATCHES "lto")
@@ -557,8 +643,14 @@ if (NOT IQTREE_FLAGS MATCHES "single")
 
     if(CLANG AND APPLE)
     	if(OSX_NATIVE_ARCHITECTURE STREQUAL "arm64")
-    		link_directories(${PROJECT_SOURCE_DIR}/libmac_m1)
-            LIST(APPEND STATIC_LIBS "${PROJECT_SOURCE_DIR}/libmac_m1/libomp.a" "${PROJECT_SOURCE_DIR}/libmac_m1/libz.a")
+            if(MACOS_MAJOR_VERSION GREATER_EQUAL 26)
+                message("Using libmac26_m1 for macOS ${MACOS_VERSION}")
+                link_directories(${PROJECT_SOURCE_DIR}/libmac26_m1)
+                LIST(APPEND STATIC_LIBS "${PROJECT_SOURCE_DIR}/libmac26_m1/libomp.a" "${PROJECT_SOURCE_DIR}/libmac_m1/libz.a")
+            else()
+                link_directories(${PROJECT_SOURCE_DIR}/libmac_m1)
+                LIST(APPEND STATIC_LIBS "${PROJECT_SOURCE_DIR}/libmac_m1/libomp.a" "${PROJECT_SOURCE_DIR}/libmac_m1/libz.a")
+            endif()
     	else()
             link_directories(${PROJECT_SOURCE_DIR}/libmac)
             LIST(APPEND STATIC_LIBS "${PROJECT_SOURCE_DIR}/libmac/libomp.a" "${PROJECT_SOURCE_DIR}/libmac/libz.a")
@@ -1007,6 +1099,11 @@ endif()
 target_link_libraries(iqtree3 pll ncl nclextra utils pda lbfgsb whtest sprng vectorclass model
     gsl alignment tree simulator terrace yaml-cpp phyloYAML main ${TARGET_CMAPLE} ${PLATFORM_LIB} ${STD_LIB} ${THREAD_LIB} ${ATOMIC_LIB})
 
+# CUDA libraries
+if (USE_CUDA)
+    target_link_libraries(iqtree3 ${CUDART_LIBRARY})
+endif()
+
 LIST(APPEND STATIC_LIBS "${PROJECT_BINARY_DIR}/pll/libpll.a" "${PROJECT_BINARY_DIR}/ncl/libncl.a" "${PROJECT_BINARY_DIR}/nclextra/libnclextra.a" "${PROJECT_BINARY_DIR}/utils/libutils.a" "${PROJECT_BINARY_DIR}/pda/libpda.a" "${PROJECT_BINARY_DIR}/lbfgsb/liblbfgsb.a" "${PROJECT_BINARY_DIR}/whtest/libwhtest.a" "${PROJECT_BINARY_DIR}/sprng/libsprng.a" "${PROJECT_BINARY_DIR}/vectorclass/libvectorclass.a" "${PROJECT_BINARY_DIR}/model/libmodel.a" "${PROJECT_BINARY_DIR}/gsl/libgsl.a" "${PROJECT_BINARY_DIR}/alignment/libalignment.a" "${PROJECT_BINARY_DIR}/tree/libtree.a" "${PROJECT_BINARY_DIR}/simulator/libsimulator.a" "${PROJECT_BINARY_DIR}/terrace/libterrace.a" "${PROJECT_BINARY_DIR}/yaml-cpp/libyaml-cpp.a" "${PROJECT_BINARY_DIR}/phylo-yaml/libphyloYAML.a" "${PROJECT_BINARY_DIR}/main/libmain.a")
 
 if (USE_CMAPLE_AA STREQUAL "ON")