ddidderr/zstd-rs
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

avx2 mode solely detected using __AVX2__ macro

Browse Source
This commit is contained in:
Yann Collet
2015-02-01 10:13:22 +01:00
parent e5c25e0ebc
commit b0802817c0
1 changed files with 7 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files
lib/zstd.c
+6 -3
View File
@@ -68,6 +68,7 @@
#include <stdio.h> /* debug : printf */ #include <stdio.h> /* debug : printf */
#include "zstd_static.h" #include "zstd_static.h"
#if defined(__clang__) || defined(__GNUC__) #if defined(__clang__) || defined(__GNUC__)
# pragma clang diagnostic ignored "-Wtypedef-redefinition"
# include "fse.c" /* due to GCC/Clang inlining limitations, including *.c runs noticeably faster */ # include "fse.c" /* due to GCC/Clang inlining limitations, including *.c runs noticeably faster */
#else #else
# include "fse_static.h" # include "fse_static.h"
@@ -77,7 +78,8 @@
/******************************************************** /********************************************************
* Compiler specifics * Compiler specifics
*********************************************************/ *********************************************************/
#if (!(defined(_MSC_VER) && (_MSC_VER<=1500))) /* exclude Visual 2008 and below */ //#if (!(defined(_MSC_VER) && (_MSC_VER<=1500))) /* exclude Visual 2008 and below */
#ifdef __AVX2__
# include <immintrin.h> /* AVX2 intrinsics */ # include <immintrin.h> /* AVX2 intrinsics */
#endif #endif
@@ -975,7 +977,7 @@ static void ZSTD_scaleDownCtx(void* ctx, const U32 limit)
U32* h = srt->hashTable; U32* h = srt->hashTable;
int i; int i;
#if defined(_INCLUDED_IMM) || defined(__AVX2__) /* <immintrin.h> */ #if defined(__AVX2__) /* <immintrin.h> */
/* AVX2 version */ /* AVX2 version */
const __m256i limit8 = _mm256_set1_epi32(limit); const __m256i limit8 = _mm256_set1_epi32(limit);
for (i=0; i<HASH_TABLESIZE; i+=8) for (i=0; i<HASH_TABLESIZE; i+=8)
@@ -1012,10 +1014,11 @@ static void ZSTD_limitCtx(void* cctx, const U32 limit)
return; return;
} }
#if defined(_INCLUDED_IMM) || defined(__AVX2__) /* <immintrin.h> */ #if defined(__AVX2__) /* <immintrin.h> */
/* AVX2 version */ /* AVX2 version */
{ {
const __m256i limit8 = _mm256_set1_epi32(limit); const __m256i limit8 = _mm256_set1_epi32(limit);
//printf("test avx2!\n");
for (i=0; i<HASH_TABLESIZE; i+=8) for (i=0; i<HASH_TABLESIZE; i+=8)
{ {
__m256i src =_mm256_loadu_si256((const __m256i*)(h+i)); __m256i src =_mm256_loadu_si256((const __m256i*)(h+i));