coffee/libyuv
1
0
Fork 0
mirror of https://chromium.googlesource.com/libyuv/libyuv synced 2026-02-11 04:39:53 +08:00
Code Issues Packages Projects Releases Wiki Activity

ColorMatrix and friends ported to NaCL

Browse Source 
BUG=253
TESTED=ncval_x86_64.exe newlib/Release/nacltest_x86_64.nexe
R=nfullagar@chromium.org

Review URL: https://webrtc-codereview.appspot.com/1981004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@747 16f28f9a-4ce2-e073-06de-1de4eb20be90
This commit is contained in:
fbarchard@google.com 2013-08-06 22:27:01 +00:00
parent abfeea9b81
commit b8ffdc9e57
5 changed files with 49 additions and 47 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: 746 Version: 747
License: BSD License: BSD
License File: LICENSE License File: LICENSE

8
include/libyuv/row.h
View File

@ -42,7 +42,11 @@ extern "C" {
#define HAS_ARGBADDROW_SSE2 #define HAS_ARGBADDROW_SSE2
#define HAS_ARGBATTENUATEROW_SSSE3 #define HAS_ARGBATTENUATEROW_SSSE3
#define HAS_ARGBBLENDROW_SSSE3 #define HAS_ARGBBLENDROW_SSSE3
#define HAS_ARGBCOLORMATRIXROW_SSSE3
#define HAS_ARGBGRAYROW_SSSE3
#define HAS_ARGBMULTIPLYROW_SSE2 #define HAS_ARGBMULTIPLYROW_SSE2
#define HAS_ARGBQUANTIZEROW_SSE2
#define HAS_ARGBSEPIAROW_SSSE3
#define HAS_ARGBSHADEROW_SSE2 #define HAS_ARGBSHADEROW_SSE2
#define HAS_ARGBSUBTRACTROW_SSE2 #define HAS_ARGBSUBTRACTROW_SSE2
@ -122,11 +126,7 @@ extern "C" {
// Effects: // Effects:
#define HAS_ARGBAFFINEROW_SSE2 #define HAS_ARGBAFFINEROW_SSE2
#define HAS_ARGBCOLORMATRIXROW_SSSE3
#define HAS_ARGBGRAYROW_SSSE3
#define HAS_ARGBMIRRORROW_SSSE3 #define HAS_ARGBMIRRORROW_SSSE3
#define HAS_ARGBQUANTIZEROW_SSE2
#define HAS_ARGBSEPIAROW_SSSE3
#define HAS_ARGBUNATTENUATEROW_SSE2 #define HAS_ARGBUNATTENUATEROW_SSE2
#define HAS_COMPUTECUMULATIVESUMROW_SSE2 #define HAS_COMPUTECUMULATIVESUMROW_SSE2
#define HAS_CUMULATIVESUMTOAVERAGEROW_SSE2 #define HAS_CUMULATIVESUMTOAVERAGEROW_SSE2

2
include/libyuv/version.h
View File

@ -11,6 +11,6 @@
#ifndef INCLUDE_LIBYUV_VERSION_H_ // NOLINT #ifndef INCLUDE_LIBYUV_VERSION_H_ // NOLINT
#define INCLUDE_LIBYUV_VERSION_H_ #define INCLUDE_LIBYUV_VERSION_H_
#define LIBYUV_VERSION 746 #define LIBYUV_VERSION 747
#endif // INCLUDE_LIBYUV_VERSION_H_ NOLINT #endif // INCLUDE_LIBYUV_VERSION_H_ NOLINT

76
source/row_posix.cc
View File

@ -21,11 +21,13 @@ extern "C" {
#if !defined(LIBYUV_DISABLE_X86) && (defined(__x86_64__) || defined(__i386__)) #if !defined(LIBYUV_DISABLE_X86) && (defined(__x86_64__) || defined(__i386__))
#if defined(__native_client__) && defined(__x86_64__) #if defined(__native_client__) && defined(__x86_64__)
#define MEMACCESS(x) "%%nacl:(%%r15,%q" #x ")" #define MEMACCESS(base) "%%nacl:(%%r15,%q" #base ")"
#define MEMLEA(x, y) #x "(%q" #y ")" #define MEMACCESS2(offset, base) "%%nacl:" #offset "(%%r15,%q" #base ")"
#define MEMLEA(offset, base) #offset "(%q" #base ")"
#else #else
#define MEMACCESS(x) "(%" #x ")" #define MEMACCESS(base) "(%" #base ")"
#define MEMLEA(x, y) #x "(%" #y ")" #define MEMACCESS2(offset, base) #offset "(%" #base ")"
#define MEMLEA(offset, base) #offset "(%" #base ")"
#endif #endif
#ifdef HAS_ARGBTOYROW_SSSE3 #ifdef HAS_ARGBTOYROW_SSSE3
@ -3925,21 +3927,21 @@ void ARGBGrayRow_SSSE3(const uint8* src_argb, uint8* dst_argb, int width) {
asm volatile ( asm volatile (
"movdqa %3,%%xmm4 \n" "movdqa %3,%%xmm4 \n"
"movdqa %4,%%xmm5 \n" "movdqa %4,%%xmm5 \n"
"sub %0,%1 \n"
// 8 pixel loop. // 8 pixel loop.
".p2align 4 \n" ".p2align 4 \n"
"1: \n" "1: \n"
"movdqa (%0),%%xmm0 \n" "movdqa "MEMACCESS(0)",%%xmm0 \n"
"movdqa 0x10(%0),%%xmm1 \n" "movdqa "MEMACCESS2(0x10,0)",%%xmm1 \n"
"pmaddubsw %%xmm4,%%xmm0 \n" "pmaddubsw %%xmm4,%%xmm0 \n"
"pmaddubsw %%xmm4,%%xmm1 \n" "pmaddubsw %%xmm4,%%xmm1 \n"
"phaddw %%xmm1,%%xmm0 \n" "phaddw %%xmm1,%%xmm0 \n"
"paddw %%xmm5,%%xmm0 \n" "paddw %%xmm5,%%xmm0 \n"
"psrlw $0x7,%%xmm0 \n" "psrlw $0x7,%%xmm0 \n"
"packuswb %%xmm0,%%xmm0 \n" "packuswb %%xmm0,%%xmm0 \n"
"movdqa (%0),%%xmm2 \n" "movdqa "MEMACCESS(0)",%%xmm2 \n"
"movdqa 0x10(%0),%%xmm3 \n" "movdqa "MEMACCESS2(0x10,0)",%%xmm3 \n"
"lea "MEMLEA(0x20,0)",%0 \n"
"psrld $0x18,%%xmm2 \n" "psrld $0x18,%%xmm2 \n"
"psrld $0x18,%%xmm3 \n" "psrld $0x18,%%xmm3 \n"
"packuswb %%xmm3,%%xmm2 \n" "packuswb %%xmm3,%%xmm2 \n"
@ -3951,9 +3953,9 @@ void ARGBGrayRow_SSSE3(const uint8* src_argb, uint8* dst_argb, int width) {
"punpcklwd %%xmm3,%%xmm0 \n" "punpcklwd %%xmm3,%%xmm0 \n"
"punpckhwd %%xmm3,%%xmm1 \n" "punpckhwd %%xmm3,%%xmm1 \n"
"sub $0x8,%2 \n" "sub $0x8,%2 \n"
"movdqa %%xmm0,(%0,%1,1) \n" "movdqa %%xmm0,"MEMACCESS(1)" \n"
"movdqa %%xmm1,0x10(%0,%1,1) \n" "movdqa %%xmm1,"MEMACCESS2(0x10,1)" \n"
"lea 0x20(%0),%0 \n" "lea "MEMLEA(0x20,1)",%1 \n"
"jg 1b \n" "jg 1b \n"
: "+r"(src_argb), // %0 : "+r"(src_argb), // %0
"+r"(dst_argb), // %1 "+r"(dst_argb), // %1
@ -3995,30 +3997,30 @@ void ARGBSepiaRow_SSSE3(uint8* dst_argb, int width) {
// 8 pixel loop. // 8 pixel loop.
".p2align 4 \n" ".p2align 4 \n"
"1: \n" "1: \n"
"movdqa (%0),%%xmm0 \n" "movdqa "MEMACCESS(0)",%%xmm0 \n"
"movdqa 0x10(%0),%%xmm6 \n" "movdqa "MEMACCESS2(0x10,0)",%%xmm6 \n"
"pmaddubsw %%xmm2,%%xmm0 \n" "pmaddubsw %%xmm2,%%xmm0 \n"
"pmaddubsw %%xmm2,%%xmm6 \n" "pmaddubsw %%xmm2,%%xmm6 \n"
"phaddw %%xmm6,%%xmm0 \n" "phaddw %%xmm6,%%xmm0 \n"
"psrlw $0x7,%%xmm0 \n" "psrlw $0x7,%%xmm0 \n"
"packuswb %%xmm0,%%xmm0 \n" "packuswb %%xmm0,%%xmm0 \n"
"movdqa (%0),%%xmm5 \n" "movdqa "MEMACCESS(0)",%%xmm5 \n"
"movdqa 0x10(%0),%%xmm1 \n" "movdqa "MEMACCESS2(0x10,0)",%%xmm1 \n"
"pmaddubsw %%xmm3,%%xmm5 \n" "pmaddubsw %%xmm3,%%xmm5 \n"
"pmaddubsw %%xmm3,%%xmm1 \n" "pmaddubsw %%xmm3,%%xmm1 \n"
"phaddw %%xmm1,%%xmm5 \n" "phaddw %%xmm1,%%xmm5 \n"
"psrlw $0x7,%%xmm5 \n" "psrlw $0x7,%%xmm5 \n"
"packuswb %%xmm5,%%xmm5 \n" "packuswb %%xmm5,%%xmm5 \n"
"punpcklbw %%xmm5,%%xmm0 \n" "punpcklbw %%xmm5,%%xmm0 \n"
"movdqa (%0),%%xmm5 \n" "movdqa "MEMACCESS(0)",%%xmm5 \n"
"movdqa 0x10(%0),%%xmm1 \n" "movdqa "MEMACCESS2(0x10,0)",%%xmm1 \n"
"pmaddubsw %%xmm4,%%xmm5 \n" "pmaddubsw %%xmm4,%%xmm5 \n"
"pmaddubsw %%xmm4,%%xmm1 \n" "pmaddubsw %%xmm4,%%xmm1 \n"
"phaddw %%xmm1,%%xmm5 \n" "phaddw %%xmm1,%%xmm5 \n"
"psrlw $0x7,%%xmm5 \n" "psrlw $0x7,%%xmm5 \n"
"packuswb %%xmm5,%%xmm5 \n" "packuswb %%xmm5,%%xmm5 \n"
"movdqa (%0),%%xmm6 \n" "movdqa "MEMACCESS(0)",%%xmm6 \n"
"movdqa 0x10(%0),%%xmm1 \n" "movdqa "MEMACCESS2(0x10,0)",%%xmm1 \n"
"psrld $0x18,%%xmm6 \n" "psrld $0x18,%%xmm6 \n"
"psrld $0x18,%%xmm1 \n" "psrld $0x18,%%xmm1 \n"
"packuswb %%xmm1,%%xmm6 \n" "packuswb %%xmm1,%%xmm6 \n"
@ -4028,9 +4030,9 @@ void ARGBSepiaRow_SSSE3(uint8* dst_argb, int width) {
"punpcklwd %%xmm5,%%xmm0 \n" "punpcklwd %%xmm5,%%xmm0 \n"
"punpckhwd %%xmm5,%%xmm1 \n" "punpckhwd %%xmm5,%%xmm1 \n"
"sub $0x8,%1 \n" "sub $0x8,%1 \n"
"movdqa %%xmm0,(%0) \n" "movdqa %%xmm0,"MEMACCESS(0)" \n"
"movdqa %%xmm1,0x10(%0) \n" "movdqa %%xmm1,"MEMACCESS2(0x10,0)" \n"
"lea 0x20(%0),%0 \n" "lea "MEMLEA(0x20,0)",%0 \n"
"jg 1b \n" "jg 1b \n"
: "+r"(dst_argb), // %0 : "+r"(dst_argb), // %0
"+r"(width) // %1 "+r"(width) // %1
@ -4061,12 +4063,12 @@ void ARGBColorMatrixRow_SSSE3(uint8* dst_argb, const int8* matrix_argb,
// 8 pixel loop. // 8 pixel loop.
".p2align 4 \n" ".p2align 4 \n"
"1: \n" "1: \n"
"movdqa (%0),%%xmm0 \n" "movdqa "MEMACCESS(0)",%%xmm0 \n"
"movdqa 0x10(%0),%%xmm6 \n" "movdqa "MEMACCESS2(0x10,0)",%%xmm6 \n"
"pmaddubsw %%xmm2,%%xmm0 \n" "pmaddubsw %%xmm2,%%xmm0 \n"
"pmaddubsw %%xmm2,%%xmm6 \n" "pmaddubsw %%xmm2,%%xmm6 \n"
"movdqa (%0),%%xmm5 \n" "movdqa "MEMACCESS(0)",%%xmm5 \n"
"movdqa 0x10(%0),%%xmm1 \n" "movdqa "MEMACCESS2(0x10,0)",%%xmm1 \n"
"pmaddubsw %%xmm3,%%xmm5 \n" "pmaddubsw %%xmm3,%%xmm5 \n"
"pmaddubsw %%xmm3,%%xmm1 \n" "pmaddubsw %%xmm3,%%xmm1 \n"
"phaddsw %%xmm6,%%xmm0 \n" "phaddsw %%xmm6,%%xmm0 \n"
@ -4076,15 +4078,15 @@ void ARGBColorMatrixRow_SSSE3(uint8* dst_argb, const int8* matrix_argb,
"packuswb %%xmm0,%%xmm0 \n" "packuswb %%xmm0,%%xmm0 \n"
"packuswb %%xmm5,%%xmm5 \n" "packuswb %%xmm5,%%xmm5 \n"
"punpcklbw %%xmm5,%%xmm0 \n" "punpcklbw %%xmm5,%%xmm0 \n"
"movdqa (%0),%%xmm5 \n" "movdqa "MEMACCESS(0)",%%xmm5 \n"
"movdqa 0x10(%0),%%xmm1 \n" "movdqa "MEMACCESS2(0x10,0)",%%xmm1 \n"
"pmaddubsw %%xmm4,%%xmm5 \n" "pmaddubsw %%xmm4,%%xmm5 \n"
"pmaddubsw %%xmm4,%%xmm1 \n" "pmaddubsw %%xmm4,%%xmm1 \n"
"phaddsw %%xmm1,%%xmm5 \n" "phaddsw %%xmm1,%%xmm5 \n"
"psraw $0x7,%%xmm5 \n" "psraw $0x7,%%xmm5 \n"
"packuswb %%xmm5,%%xmm5 \n" "packuswb %%xmm5,%%xmm5 \n"
"movdqa (%0),%%xmm6 \n" "movdqa "MEMACCESS(0)",%%xmm6 \n"
"movdqa 0x10(%0),%%xmm1 \n" "movdqa "MEMACCESS2(0x10,0)",%%xmm1 \n"
"psrld $0x18,%%xmm6 \n" "psrld $0x18,%%xmm6 \n"
"psrld $0x18,%%xmm1 \n" "psrld $0x18,%%xmm1 \n"
"packuswb %%xmm1,%%xmm6 \n" "packuswb %%xmm1,%%xmm6 \n"
@ -4094,9 +4096,9 @@ void ARGBColorMatrixRow_SSSE3(uint8* dst_argb, const int8* matrix_argb,
"punpcklwd %%xmm5,%%xmm0 \n" "punpcklwd %%xmm5,%%xmm0 \n"
"punpckhwd %%xmm5,%%xmm1 \n" "punpckhwd %%xmm5,%%xmm1 \n"
"sub $0x8,%1 \n" "sub $0x8,%1 \n"
"movdqa %%xmm0,(%0) \n" "movdqa %%xmm0,"MEMACCESS(0)" \n"
"movdqa %%xmm1,0x10(%0) \n" "movdqa %%xmm1,"MEMACCESS2(0x10,0)" \n"
"lea 0x20(%0),%0 \n" "lea "MEMLEA(0x20,0)",%0 \n"
"jg 1b \n" "jg 1b \n"
: "+r"(dst_argb), // %0 : "+r"(dst_argb), // %0
"+r"(width) // %1 "+r"(width) // %1
@ -4131,7 +4133,7 @@ void ARGBQuantizeRow_SSE2(uint8* dst_argb, int scale, int interval_size,
// 4 pixel loop. // 4 pixel loop.
".p2align 2 \n" ".p2align 2 \n"
"1: \n" "1: \n"
"movdqa (%0),%%xmm0 \n" "movdqa "MEMACCESS(0)",%%xmm0 \n"
"punpcklbw %%xmm5,%%xmm0 \n" "punpcklbw %%xmm5,%%xmm0 \n"
"pmulhuw %%xmm2,%%xmm0 \n" "pmulhuw %%xmm2,%%xmm0 \n"
"movdqa (%0),%%xmm1 \n" "movdqa (%0),%%xmm1 \n"
@ -4146,8 +4148,8 @@ void ARGBQuantizeRow_SSE2(uint8* dst_argb, int scale, int interval_size,
"packuswb %%xmm1,%%xmm0 \n" "packuswb %%xmm1,%%xmm0 \n"
"por %%xmm7,%%xmm0 \n" "por %%xmm7,%%xmm0 \n"
"sub $0x4,%1 \n" "sub $0x4,%1 \n"
"movdqa %%xmm0,(%0) \n" "movdqa %%xmm0,"MEMACCESS(0)" \n"
"lea 0x10(%0),%0 \n" "lea "MEMLEA(0x10,0)",%0 \n"
"jg 1b \n" "jg 1b \n"
: "+r"(dst_argb), // %0 : "+r"(dst_argb), // %0
"+r"(width) // %1 "+r"(width) // %1

8
source/row_win.cc
View File

@ -4922,7 +4922,6 @@ void ARGBGrayRow_SSSE3(const uint8* src_argb, uint8* dst_argb, int width) {
mov ecx, [esp + 12] /* width */ mov ecx, [esp + 12] /* width */
movdqa xmm4, kARGBToYJ movdqa xmm4, kARGBToYJ
movdqa xmm5, kAddYJ64 movdqa xmm5, kAddYJ64
sub edx, eax
align 16 align 16
convertloop: convertloop:
@ -4936,6 +4935,7 @@ void ARGBGrayRow_SSSE3(const uint8* src_argb, uint8* dst_argb, int width) {
packuswb xmm0, xmm0 // 8 G bytes packuswb xmm0, xmm0 // 8 G bytes
movdqa xmm2, [eax] // A movdqa xmm2, [eax] // A
movdqa xmm3, [eax + 16] movdqa xmm3, [eax + 16]
lea eax, [eax + 32]
psrld xmm2, 24 psrld xmm2, 24
psrld xmm3, 24 psrld xmm3, 24
packuswb xmm2, xmm3 packuswb xmm2, xmm3
@ -4947,9 +4947,9 @@ void ARGBGrayRow_SSSE3(const uint8* src_argb, uint8* dst_argb, int width) {
punpcklwd xmm0, xmm3 // GGGA first 4 punpcklwd xmm0, xmm3 // GGGA first 4
punpckhwd xmm1, xmm3 // GGGA next 4 punpckhwd xmm1, xmm3 // GGGA next 4
sub ecx, 8 sub ecx, 8
movdqa [eax + edx], xmm0 movdqa [edx], xmm0
movdqa [eax + edx + 16], xmm1 movdqa [edx + 16], xmm1
lea eax, [eax + 32] lea edx, [edx + 32]
jg convertloop jg convertloop
ret ret
} }