Skip to content

SSE2 implementation for LzFind_SaturSub() #140

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
yumeyao wants to merge 1 commit into
base: main
Choose a base branch
from
Open

SSE2 implementation for LzFind_SaturSub() #140

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
31 changes: 30 additions & 1 deletion C/LzFind.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -598,7 +598,7 @@ void MatchFinder_Init(void *_p)

#ifdef MY_CPU_X86_OR_AMD64
#if defined(__clang__) && (__clang_major__ >= 4) \
|| defined(Z7_GCC_VERSION) && (Z7_GCC_VERSION >= 40701)
|| defined(Z7_GCC_VERSION) && (Z7_GCC_VERSION >= 40900)
// || defined(__INTEL_COMPILER) && (__INTEL_COMPILER >= 1900)

#define USE_LZFIND_SATUR_SUB_128
Expand All @@ -614,6 +614,12 @@ void MatchFinder_Init(void *_p)
#endif
#endif

#ifdef MY_CPU_SSE2
// x86[-64] doesn't have integer satur ops instructions (thus slow if not vectorized)
// Since all x86-64 CPUs support SSE2, we can use SSE2 as a baseline implementation.
#include <immintrin.h>
#endif

#elif defined(MY_CPU_ARM64) \
/* || (defined(__ARM_ARCH) && (__ARM_ARCH >= 7)) */

Expand Down Expand Up @@ -777,6 +783,28 @@ void
Z7_FASTCALL
LzFind_SaturSub_32(UInt32 subValue, CLzRef *items, const CLzRef *lim)
{
// This SSE2 implementation is taken from clang >=11 opt result.
// GCC can't do it in a clean way. MSVC can't even vectorize it.
#ifdef MY_CPU_SSE2
const __m128i sub2 = _mm_set_epi32((Int32)subValue, (Int32)subValue, (Int32)subValue, (Int32)subValue);
const __m128i sign2 = _mm_set_epi32(0x80000000, 0x80000000, 0x80000000, 0x80000000);
const __m128i cmp2 = _mm_xor_si128(sub2, sign2);

// by reversing the sign (adding 0x80000000), we flip the strict order of signed/unsigned number as well.
#define SASUB_SSE2_RSIGN(v) _mm_xor_si128(v, sign2)
// so we can use signed compare on unsigned numbers with their sign reversed.
#define SASUB_SSE2_CMPGT(v) _mm_cmpgt_epi32(SASUB_SSE2_RSIGN(v), cmp2)
// the mask can be used to select like (c ? a : b), but here since b is 0, then we just use (a & c)
#define SASUB_SSE2_v(v) _mm_and_si128(_mm_sub_epi32(v, sub2), SASUB_SSE2_CMPGT(v))
#define SASUB_SSE2(i) *(__m128i *)(void *)(items + (i) * 4) = SASUB_SSE2_v(*(const __m128i *)(const void *)(items + (i) * 4));
Z7_PRAGMA_OPT_DISABLE_LOOP_UNROLL_VECTORIZE
do
{
SASUB_SSE2(0) SASUB_SSE2(1) items += 2 * 4;
SASUB_SSE2(0) SASUB_SSE2(1) items += 2 * 4;
}
while (items != lim);
#else
Z7_PRAGMA_OPT_DISABLE_LOOP_UNROLL_VECTORIZE
do
{
Expand All @@ -786,6 +814,7 @@ LzFind_SaturSub_32(UInt32 subValue, CLzRef *items, const CLzRef *lim)
SASUB_32(0) SASUB_32(1) items += 2;
}
while (items != lim);
#endif
}

#endif
Expand Down