coffee/libyuv
1
0
Fork 0
mirror of https://chromium.googlesource.com/libyuv/libyuv synced 2025-12-06 16:56:55 +08:00
Code Issues Packages Projects Releases Wiki Activity

CopyRow_X86 for gcc

Browse Source 
BUG=none
TEST=none
Review URL: http://webrtc-codereview.appspot.com/300007

git-svn-id: http://libyuv.googlecode.com/svn/trunk@97 16f28f9a-4ce2-e073-06de-1de4eb20be90
This commit is contained in:
fbarchard@google.com 2011-12-05 22:45:32 +00:00
parent c2b74366ec
commit e5ffa14f31
2 changed files with 26 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
README.chromium
View File

@ -1,6 +1,6 @@
Name: libyuv Name: libyuv
URL: http://code.google.com/p/libyuv/ URL: http://code.google.com/p/libyuv/
Version: 96 Version: 97
License: BSD License: BSD
License File: LICENSE License File: LICENSE

36
source/planar_functions.cc
View File

@ -129,7 +129,6 @@ static void SplitUV_C(const uint8* src_uv,
// CopyRows copys 'count' bytes using a 16 byte load/store, 64 bytes at time // CopyRows copys 'count' bytes using a 16 byte load/store, 64 bytes at time
#if defined(_M_IX86) && !defined(YUV_DISABLE_ASM) #if defined(_M_IX86) && !defined(YUV_DISABLE_ASM)
#define HAS_COPYROW_SSE2 #define HAS_COPYROW_SSE2
#define HAS_COPYROW_X86
__declspec(naked) __declspec(naked)
void CopyRow_SSE2(const uint8* src, uint8* dst, int count) { void CopyRow_SSE2(const uint8* src, uint8* dst, int count) {
__asm { __asm {
@ -150,6 +149,7 @@ void CopyRow_SSE2(const uint8* src, uint8* dst, int count) {
} }
} }
#define HAS_COPYROW_X86
__declspec(naked) __declspec(naked)
void CopyRow_X86(const uint8* src, uint8* dst, int count) { void CopyRow_X86(const uint8* src, uint8* dst, int count) {
__asm { __asm {
@ -169,15 +169,15 @@ void CopyRow_X86(const uint8* src, uint8* dst, int count) {
#define HAS_COPYROW_SSE2 #define HAS_COPYROW_SSE2
void CopyRow_SSE2(const uint8* src, uint8* dst, int count) { void CopyRow_SSE2(const uint8* src, uint8* dst, int count) {
asm volatile ( asm volatile (
"1: \n" "1: \n"
"movdqa (%0),%%xmm0 \n" "movdqa (%0),%%xmm0 \n"
"movdqa 0x10(%0),%%xmm1 \n" "movdqa 0x10(%0),%%xmm1 \n"
"lea 0x20(%0),%0 \n" "lea 0x20(%0),%0 \n"
"movdqa %%xmm0,(%1) \n" "movdqa %%xmm0,(%1) \n"
"movdqa %%xmm1,0x10(%1) \n" "movdqa %%xmm1,0x10(%1) \n"
"lea 0x20(%1),%1 \n" "lea 0x20(%1),%1 \n"
"sub $0x20,%2 \n" "sub $0x20,%2 \n"
"ja 1b \n" "ja 1b \n"
: "+r"(src), // %0 : "+r"(src), // %0
"+r"(dst), // %1 "+r"(dst), // %1
"+r"(count) // %2 "+r"(count) // %2
@ -186,7 +186,21 @@ void CopyRow_SSE2(const uint8* src, uint8* dst, int count) {
#if defined(__SSE2__) #if defined(__SSE2__)
, "xmm0", "xmm1" , "xmm0", "xmm1"
#endif #endif
); );
}
#define HAS_COPYROW_X86
void CopyRow_X86(const uint8* src, uint8* dst, int width) {
size_t width_tmp = static_cast<size_t>(width);
asm volatile (
"shr $0x2,%2 \n"
"rep movsl (%0),(%1) \n"
: "+S"(src), // %0
"+D"(dst), // %1
"+c"(width_tmp) // %2
:
: "memory", "cc"
);
} }
#endif #endif