ligang/vpp
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

vppinfra: revert clib_memcpy optimization

Browse Source 
Looks like some compiler versions are producing wrong code when we are
copying 9-16 bytes so reverting back to the original code.

Change-Id: I74b5fa54a3b01f6288648f1cb0926030edd3b26f
Signed-off-by: Damjan Marion <damarion@cisco.com>
This commit is contained in:
Damjan Marion
2017-07-05 18:15:08 +02:00
parent 02989064e4
commit 331f66a5b4
2 changed files with 14 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
src/vppinfra/memcpy_avx.h
View File

@ -51,19 +51,22 @@
#include <stdint.h>
#include <x86intrin.h>
typedef u8 u8x16u __attribute__ ((vector_size (16), aligned (1)));
typedef u8 u8x32u __attribute__ ((vector_size (32), aligned (1)));
static inline void
clib_mov16 (u8 * dst, const u8 * src)
{
*(u8x16u *) dst = *(u8x16u *) src;
__m128i xmm0;
xmm0 = _mm_loadu_si128 ((const __m128i *) src);
_mm_storeu_si128 ((__m128i *) dst, xmm0);
}
static inline void
clib_mov32 (u8 * dst, const u8 * src)
{
*(u8x32u *) dst = *(u8x32u *) src;
__m256i ymm0;
ymm0 = _mm256_loadu_si256 ((const __m256i *) src);
_mm256_storeu_si256 ((__m256i *) dst, ymm0);
}
static inline void

11
src/vppinfra/memcpy_sse3.h
View File

@ -51,19 +51,20 @@
#include <stdint.h>
#include <x86intrin.h>
typedef u8 u8x16u __attribute__ ((vector_size (16), aligned (1)));
typedef u8 u8x32u __attribute__ ((vector_size (32), aligned (1)));
static inline void
clib_mov16 (u8 * dst, const u8 * src)
{
*(u8x16u *) dst = *(u8x16u *) src;
__m128i xmm0;
xmm0 = _mm_loadu_si128 ((const __m128i *) src);
_mm_storeu_si128 ((__m128i *) dst, xmm0);
}
static inline void
clib_mov32 (u8 * dst, const u8 * src)
{
*(u8x32u *) dst = *(u8x32u *) src;
clib_mov16 ((u8 *) dst + 0 * 16, (const u8 *) src + 0 * 16);
clib_mov16 ((u8 *) dst + 1 * 16, (const u8 *) src + 1 * 16);
}
static inline void