Skip to content

L
libyuv
Commit 540e8af8 authored by fbarchard@google.com's avatar fbarchard@google.com
Browse files
Options

remove add 16 from ARGBToYJ and add rounding, for consistency with Windows… 

remove add 16 from ARGBToYJ and add rounding, for consistency with Windows version.  row.h header macros sorted alphabetically.
BUG=269
TESTED=untested
R=tpsiaki@google.com

Review URL: https://webrtc-codereview.appspot.com/32579005

git-svn-id: http://libyuv.googlecode.com/svn/trunk@1185 16f28f9a-4ce2-e073-06de-1de4eb20be90
parent b036cf70
Hide whitespace changes
Inline Side-by-side
Showing with 149 additions and 139 deletions
README.chromium
View file @ 540e8af8
Name: libyuv
URL: http://code.google.com/p/libyuv/
Version: 1184
Version: 1185
License: BSD
License File: LICENSE
......
include/libyuv/row.h
View file @ 540e8af8
......@@ -57,38 +57,7 @@ extern "C" {
// The following are available on all x86 platforms:
#if !defined(LIBYUV_DISABLE_X86) && \
(defined(_M_IX86) || defined(__x86_64__) || defined(__i386__))
// Effects:
#define HAS_ARGBADDROW_SSE2
#define HAS_ARGBAFFINEROW_SSE2
#define HAS_ARGBATTENUATEROW_SSSE3
#define HAS_ARGBBLENDROW_SSSE3
#define HAS_ARGBCOLORMATRIXROW_SSSE3
#define HAS_ARGBCOLORTABLEROW_X86
#define HAS_ARGBCOPYALPHAROW_SSE2
#define HAS_ARGBCOPYYTOALPHAROW_SSE2
#define HAS_ARGBGRAYROW_SSSE3
#define HAS_ARGBLUMACOLORTABLEROW_SSSE3
#define HAS_ARGBMIRRORROW_SSE2
#define HAS_ARGBMULTIPLYROW_SSE2
#define HAS_ARGBPOLYNOMIALROW_SSE2
#define HAS_ARGBQUANTIZEROW_SSE2
#define HAS_ARGBSEPIAROW_SSSE3
#define HAS_ARGBSHADEROW_SSE2
#define HAS_ARGBSUBTRACTROW_SSE2
#define HAS_ARGBUNATTENUATEROW_SSE2
#define HAS_COMPUTECUMULATIVESUMROW_SSE2
#define HAS_CUMULATIVESUMTOAVERAGEROW_SSE2
#define HAS_INTERPOLATEROW_SSE2
#define HAS_INTERPOLATEROW_SSSE3
#define HAS_RGBCOLORTABLEROW_X86
#define HAS_SOBELROW_SSE2
#define HAS_SOBELTOPLANEROW_SSE2
#define HAS_SOBELXROW_SSE2
#define HAS_SOBELXYROW_SSE2
#define HAS_SOBELYROW_SSE2
// Conversions:
#define HAS_ARGBTOUVROW_SSSE3
#define HAS_ABGRTOUVROW_SSSE3
#define HAS_ABGRTOYROW_SSSE3
#define HAS_ARGB1555TOARGBROW_SSE2
......@@ -105,6 +74,7 @@ extern "C" {
#define HAS_ARGBTOUV422ROW_SSSE3
#define HAS_ARGBTOUV444ROW_SSSE3
#define HAS_ARGBTOUVJROW_SSSE3
#define HAS_ARGBTOUVROW_SSSE3
#define HAS_ARGBTOYJROW_SSSE3
#define HAS_ARGBTOYROW_SSSE3
#define HAS_BGRATOUVROW_SSSE3
......@@ -113,9 +83,9 @@ extern "C" {
#define HAS_COPYROW_SSE2
#define HAS_I400TOARGBROW_SSE2
#define HAS_I411TOARGBROW_SSSE3
#define HAS_I422TOARGB1555ROW_SSSE3
#define HAS_I422TOABGRROW_SSSE3
#define HAS_I422TOARGB1555ROW_SSSE3
#define HAS_I422TOARGB1555ROW_SSSE3
#define HAS_I422TOARGB4444ROW_SSSE3
#define HAS_I422TOARGBROW_SSSE3
#define HAS_I422TOBGRAROW_SSSE3
......@@ -153,6 +123,36 @@ extern "C" {
#define HAS_YUY2TOUV422ROW_SSE2
#define HAS_YUY2TOUVROW_SSE2
#define HAS_YUY2TOYROW_SSE2
// Effects:
#define HAS_ARGBADDROW_SSE2
#define HAS_ARGBAFFINEROW_SSE2
#define HAS_ARGBATTENUATEROW_SSSE3
#define HAS_ARGBBLENDROW_SSSE3
#define HAS_ARGBCOLORMATRIXROW_SSSE3
#define HAS_ARGBCOLORTABLEROW_X86
#define HAS_ARGBCOPYALPHAROW_SSE2
#define HAS_ARGBCOPYYTOALPHAROW_SSE2
#define HAS_ARGBGRAYROW_SSSE3
#define HAS_ARGBLUMACOLORTABLEROW_SSSE3
#define HAS_ARGBMIRRORROW_SSE2
#define HAS_ARGBMULTIPLYROW_SSE2
#define HAS_ARGBPOLYNOMIALROW_SSE2
#define HAS_ARGBQUANTIZEROW_SSE2
#define HAS_ARGBSEPIAROW_SSSE3
#define HAS_ARGBSHADEROW_SSE2
#define HAS_ARGBSUBTRACTROW_SSE2
#define HAS_ARGBUNATTENUATEROW_SSE2
#define HAS_COMPUTECUMULATIVESUMROW_SSE2
#define HAS_CUMULATIVESUMTOAVERAGEROW_SSE2
#define HAS_INTERPOLATEROW_SSE2
#define HAS_INTERPOLATEROW_SSSE3
#define HAS_RGBCOLORTABLEROW_X86
#define HAS_SOBELROW_SSE2
#define HAS_SOBELTOPLANEROW_SSE2
#define HAS_SOBELXROW_SSE2
#define HAS_SOBELXYROW_SSE2
#define HAS_SOBELYROW_SSE2
#endif
// The following are available on x64 Visual C:
......@@ -184,33 +184,33 @@ extern "C" {
// The code supports NaCL but requires a new compiler and validator.
#if !defined(LIBYUV_DISABLE_X86) && (defined(VISUALC_HAS_AVX2) || \
defined(CLANG_HAS_AVX2) || defined(GCC_HAS_AVX2))
#define HAS_COPYROW_AVX
#define HAS_ARGBPOLYNOMIALROW_AVX2
#define HAS_ARGBSHUFFLEROW_AVX2
#define HAS_ARGBCOPYALPHAROW_AVX2
#define HAS_ARGBCOPYYTOALPHAROW_AVX2
#define HAS_ARGBMIRRORROW_AVX2
#define HAS_ARGBPOLYNOMIALROW_AVX2
#define HAS_ARGBSHUFFLEROW_AVX2
#define HAS_ARGBTOYJROW_AVX2
#define HAS_ARGBTOYROW_AVX2
#define HAS_COPYROW_AVX
#define HAS_I422TOABGRROW_AVX2
#define HAS_I422TOARGBROW_AVX2
#define HAS_I422TOBGRAROW_AVX2
#define HAS_I422TOABGRROW_AVX2
#define HAS_I422TORGBAROW_AVX2
#define HAS_YUY2TOYROW_AVX2
#define HAS_YUY2TOUV422ROW_AVX2
#define HAS_YUY2TOUVROW_AVX2
#define HAS_UYVYTOYROW_AVX2
#define HAS_UYVYTOUV422ROW_AVX2
#define HAS_UYVYTOUVROW_AVX2
#define HAS_SPLITUVROW_AVX2
#define HAS_MERGEUVROW_AVX2
#define HAS_MIRRORROW_AVX2
#define HAS_ARGBMIRRORROW_AVX2
#define HAS_ARGBTOYROW_AVX2
#define HAS_ARGBTOYJROW_AVX2
#define HAS_SPLITUVROW_AVX2
#define HAS_UYVYTOUV422ROW_AVX2
#define HAS_UYVYTOUVROW_AVX2
#define HAS_UYVYTOYROW_AVX2
#define HAS_YUY2TOUV422ROW_AVX2
#define HAS_YUY2TOUVROW_AVX2
#define HAS_YUY2TOYROW_AVX2
// Effects:
#define HAS_ARGBADDROW_AVX2
#define HAS_ARGBSUBTRACTROW_AVX2
#define HAS_ARGBMULTIPLYROW_AVX2
#define HAS_ARGBATTENUATEROW_AVX2
#define HAS_ARGBMULTIPLYROW_AVX2
#define HAS_ARGBSUBTRACTROW_AVX2
#define HAS_ARGBUNATTENUATEROW_AVX2
#endif
......@@ -240,100 +240,101 @@ extern "C" {
#if !defined(LIBYUV_DISABLE_X86) && \
(defined(_M_IX86) || defined(__x86_64__) || defined(__i386__)) && \
!defined(LIBYUV_SSSE3_ONLY)
#define HAS_ARGBBLENDROW_SSE2
#define HAS_ARGBATTENUATEROW_SSE2
#define HAS_ARGBBLENDROW_SSE2
#define HAS_MIRRORROW_SSE2
#endif
// The following are available on arm64 platforms:
// TODO(fbarchard): Merge arm64 and arm7 neon macros.
#if !defined(LIBYUV_DISABLE_NEON) && defined(__aarch64__)
#define HAS_I444TOARGBROW_NEON
#define HAS_I422TOARGBROW_NEON
#define HAS_I411TOARGBROW_NEON
#define HAS_I422TOBGRAROW_NEON
#define HAS_I422TOABGRROW_NEON
#define HAS_I422TORGBAROW_NEON
#define HAS_I422TORGB24ROW_NEON
#define HAS_I422TORAWROW_NEON
#define HAS_I422TORGB565ROW_NEON
#define HAS_I422TOARGB1555ROW_NEON
#define HAS_I422TOARGB4444ROW_NEON
#define HAS_YTOARGBROW_NEON
#define HAS_I400TOARGBROW_NEON
#define HAS_NV12TOARGBROW_NEON
#define HAS_NV21TOARGBROW_NEON
#define HAS_NV12TORGB565ROW_NEON
#define HAS_NV21TORGB565ROW_NEON
#define HAS_YUY2TOARGBROW_NEON
#define HAS_UYVYTOARGBROW_NEON
#define HAS_SPLITUVROW_NEON
#define HAS_MERGEUVROW_NEON
#define HAS_COPYROW_NEON
#define HAS_SETROW_NEON
#define HAS_ARGBSETROWS_NEON
#define HAS_MIRRORROW_NEON
#define HAS_MIRRORUVROW_NEON
#define HAS_ARGBMIRRORROW_NEON
#define HAS_RGB24TOARGBROW_NEON
#define HAS_RAWTOARGBROW_NEON
#define HAS_RGB565TOARGBROW_NEON
#define HAS_ABGRTOUVROW_NEON
#define HAS_ABGRTOYROW_NEON
#define HAS_ARGB1555TOARGBROW_NEON
#define HAS_ARGB1555TOUVROW_NEON
#define HAS_ARGB1555TOYROW_NEON
#define HAS_ARGB4444TOARGBROW_NEON
#define HAS_ARGBTORGB24ROW_NEON
#define HAS_ARGBTORAWROW_NEON
#define HAS_YUY2TOYROW_NEON
#define HAS_UYVYTOYROW_NEON
#define HAS_YUY2TOUV422ROW_NEON
#define HAS_UYVYTOUV422ROW_NEON
#define HAS_YUY2TOUVROW_NEON
#define HAS_UYVYTOUVROW_NEON
#define HAS_ARGBTOBAYERROW_NEON
#define HAS_ARGBTOBAYERGGROW_NEON
#define HAS_ARGB4444TOUVROW_NEON
#define HAS_ARGB4444TOYROW_NEON
#define HAS_ARGBADDROW_NEON
#define HAS_ARGBATTENUATEROW_NEON
#define HAS_ARGBBLENDROW_NEON
#define HAS_ARGBCOLORMATRIXROW_NEON
#define HAS_ARGBGRAYROW_NEON
#define HAS_ARGBMIRRORROW_NEON
#define HAS_ARGBMULTIPLYROW_NEON
#define HAS_ARGBQUANTIZEROW_NEON
#define HAS_ARGBSEPIAROW_NEON
#define HAS_ARGBSETROWS_NEON
#define HAS_ARGBSHADEROW_NEON
#define HAS_ARGBSHUFFLEROW_NEON
#define HAS_I422TOYUY2ROW_NEON
#define HAS_I422TOUYVYROW_NEON
#define HAS_ARGBTORGB565ROW_NEON
#define HAS_ARGBSUBTRACTROW_NEON
#define HAS_ARGBTOARGB1555ROW_NEON
#define HAS_ARGBTOARGB4444ROW_NEON
#define HAS_ARGBTOYROW_NEON
#define HAS_ARGBTOYJROW_NEON
#define HAS_ARGBTOUV444ROW_NEON
#define HAS_ARGBTOUV422ROW_NEON
#define HAS_ARGBTOBAYERGGROW_NEON
#define HAS_ARGBTOBAYERROW_NEON
#define HAS_ARGBTORAWROW_NEON
#define HAS_ARGBTORGB24ROW_NEON
#define HAS_ARGBTORGB565ROW_NEON
#define HAS_ARGBTOUV411ROW_NEON
#define HAS_ARGBTOUVROW_NEON
#define HAS_ARGBTOUV422ROW_NEON
#define HAS_ARGBTOUV444ROW_NEON
#define HAS_ARGBTOUVJROW_NEON
#define HAS_ARGBTOUVROW_NEON
#define HAS_ARGBTOYJROW_NEON
#define HAS_ARGBTOYROW_NEON
#define HAS_BGRATOUVROW_NEON
#define HAS_ABGRTOUVROW_NEON
#define HAS_RGBATOUVROW_NEON
#define HAS_RGB24TOUVROW_NEON
#define HAS_BGRATOYROW_NEON
#define HAS_COPYROW_NEON
#define HAS_I400TOARGBROW_NEON
#define HAS_I411TOARGBROW_NEON
#define HAS_I422TOABGRROW_NEON
#define HAS_I422TOARGB1555ROW_NEON
#define HAS_I422TOARGB4444ROW_NEON
#define HAS_I422TOARGBROW_NEON
#define HAS_I422TOBGRAROW_NEON
#define HAS_I422TORAWROW_NEON
#define HAS_I422TORGB24ROW_NEON
#define HAS_I422TORGB565ROW_NEON
#define HAS_I422TORGBAROW_NEON
#define HAS_I422TOUYVYROW_NEON
#define HAS_I422TOYUY2ROW_NEON
#define HAS_I444TOARGBROW_NEON
#define HAS_INTERPOLATEROW_NEON
#define HAS_MERGEUVROW_NEON
#define HAS_MIRRORROW_NEON
#define HAS_MIRRORUVROW_NEON
#define HAS_NV12TOARGBROW_NEON
#define HAS_NV12TORGB565ROW_NEON
#define HAS_NV21TOARGBROW_NEON
#define HAS_NV21TORGB565ROW_NEON
#define HAS_RAWTOARGBROW_NEON
#define HAS_RAWTOUVROW_NEON
#define HAS_RAWTOYROW_NEON
#define HAS_RGB24TOARGBROW_NEON
#define HAS_RGB24TOUVROW_NEON
#define HAS_RGB24TOYROW_NEON
#define HAS_RGB565TOARGBROW_NEON
#define HAS_RGB565TOUVROW_NEON
#define HAS_ARGB1555TOUVROW_NEON
#define HAS_ARGB4444TOUVROW_NEON
#define HAS_RGB565TOYROW_NEON
#define HAS_ARGB1555TOYROW_NEON
#define HAS_ARGB4444TOYROW_NEON
#define HAS_BGRATOYROW_NEON
#define HAS_ABGRTOYROW_NEON
#define HAS_RGBATOUVROW_NEON
#define HAS_RGBATOYROW_NEON
#define HAS_RGB24TOYROW_NEON
#define HAS_RAWTOYROW_NEON
#define HAS_INTERPOLATEROW_NEON
#define HAS_ARGBBLENDROW_NEON
#define HAS_ARGBATTENUATEROW_NEON
#define HAS_ARGBQUANTIZEROW_NEON
#define HAS_ARGBSHADEROW_NEON
#define HAS_ARGBGRAYROW_NEON
#define HAS_ARGBSEPIAROW_NEON
#define HAS_ARGBCOLORMATRIXROW_NEON
#define HAS_ARGBMULTIPLYROW_NEON
#define HAS_ARGBADDROW_NEON
#define HAS_ARGBSUBTRACTROW_NEON
#define HAS_SETROW_NEON
#define HAS_SOBELROW_NEON
#define HAS_SOBELTOPLANEROW_NEON
#define HAS_SOBELXYROW_NEON
#define HAS_SOBELXROW_NEON
#define HAS_SOBELXYROW_NEON
#define HAS_SOBELYROW_NEON
#define HAS_SPLITUVROW_NEON
#define HAS_UYVYTOARGBROW_NEON
#define HAS_UYVYTOUV422ROW_NEON
#define HAS_UYVYTOUVROW_NEON
#define HAS_UYVYTOYROW_NEON
#define HAS_YTOARGBROW_NEON
#define HAS_YUY2TOARGBROW_NEON
#define HAS_YUY2TOUV422ROW_NEON
#define HAS_YUY2TOUVROW_NEON
#define HAS_YUY2TOYROW_NEON
#endif
// The following are available on Neon platforms:
......@@ -349,18 +350,18 @@ extern "C" {
#define HAS_ARGB4444TOYROW_NEON
#define HAS_ARGBTOARGB1555ROW_NEON
#define HAS_ARGBTOARGB4444ROW_NEON
#define HAS_ARGBTOBAYERROW_NEON
#define HAS_ARGBTOBAYERGGROW_NEON
#define HAS_ARGBTOBAYERROW_NEON
#define HAS_ARGBTORAWROW_NEON
#define HAS_ARGBTORGB24ROW_NEON
#define HAS_ARGBTORGB565ROW_NEON
#define HAS_ARGBTOUV411ROW_NEON
#define HAS_ARGBTOUV422ROW_NEON
#define HAS_ARGBTOUV444ROW_NEON
#define HAS_ARGBTOUVROW_NEON
#define HAS_ARGBTOUVJROW_NEON
#define HAS_ARGBTOYROW_NEON
#define HAS_ARGBTOUVROW_NEON
#define HAS_ARGBTOYJROW_NEON
#define HAS_ARGBTOYROW_NEON
#define HAS_BGRATOUVROW_NEON
#define HAS_BGRATOYROW_NEON
#define HAS_COPYROW_NEON
......@@ -419,12 +420,12 @@ extern "C" {
#define HAS_ARGBSEPIAROW_NEON
#define HAS_ARGBSHADEROW_NEON
#define HAS_ARGBSUBTRACTROW_NEON
#define HAS_INTERPOLATEROW_NEON
#define HAS_SOBELROW_NEON
#define HAS_SOBELTOPLANEROW_NEON
#define HAS_SOBELXYROW_NEON
#define HAS_SOBELXROW_NEON
#define HAS_SOBELXYROW_NEON
#define HAS_SOBELYROW_NEON
#define HAS_INTERPOLATEROW_NEON
// TODO(fbarchard): Investigate neon unittest failure.
// #define HAS_ARGBCOLORMATRIXROW_NEON
#endif
......
include/libyuv/version.h
View file @ 540e8af8
......@@ -11,6 +11,6 @@
#ifndef INCLUDE_LIBYUV_VERSION_H_ // NOLINT
#define INCLUDE_LIBYUV_VERSION_H_
#define LIBYUV_VERSION 1184
#define LIBYUV_VERSION 1185
#endif // INCLUDE_LIBYUV_VERSION_H_ NOLINT
source/row_posix.cc
View file @ 540e8af8
......@@ -92,6 +92,7 @@ static uvec8 kAddY16 = {
16u, 16u, 16u, 16u, 16u, 16u, 16u, 16u, 16u, 16u, 16u, 16u, 16u, 16u, 16u, 16u
};
// 7 bit fixed point 0.5.
static vec16 kAddYJ64 = {
64, 64, 64, 64, 64, 64, 64, 64
};
......@@ -704,6 +705,7 @@ void ARGBToARGB4444Row_SSE2(const uint8* src, uint8* dst, int pix) {
#endif // HAS_RGB24TOARGBROW_SSSE3
#ifdef HAS_ARGBTOYROW_SSSE3
// Convert 16 ARGB pixels (64 bytes) to 16 Y values.
void ARGBToYRow_SSSE3(const uint8* src_argb, uint8* dst_y, int pix) {
asm volatile (
"movdqa %3,%%xmm4 \n"
......@@ -743,6 +745,8 @@ void ARGBToYRow_SSSE3(const uint8* src_argb, uint8* dst_y, int pix) {
#endif // HAS_ARGBTOYROW_SSSE3
#ifdef HAS_ARGBTOYJROW_SSSE3
// Convert 16 ARGB pixels (64 bytes) to 16 YJ values.
// Same as ARGBToYRow but different coefficients, no add 16, but do rounding.
void ARGBToYJRow_SSSE3(const uint8* src_argb, uint8* dst_y, int pix) {
asm volatile (
"movdqa %3,%%xmm4 \n"
......@@ -788,6 +792,7 @@ static const lvec32 kPermdARGBToY_AVX = {
0, 4, 1, 5, 2, 6, 3, 7
};
// Convert 32 ARGB pixels (128 bytes) to 32 Y values.
void ARGBToYRow_AVX2(const uint8* src_argb, uint8* dst_y, int pix) {
asm volatile (
"vbroadcastf128 %3,%%ymm4 \n"
......@@ -804,13 +809,13 @@ void ARGBToYRow_AVX2(const uint8* src_argb, uint8* dst_y, int pix) {
"vpmaddubsw %%ymm4,%%ymm2,%%ymm2 \n"
"vpmaddubsw %%ymm4,%%ymm3,%%ymm3 \n"
"lea " MEMLEA(0x80,0) ",%0 \n"
"vphaddw %%ymm1,%%ymm0,%%ymm0 \n"
"vphaddw %%ymm1,%%ymm0,%%ymm0 \n" // mutates.
"vphaddw %%ymm3,%%ymm2,%%ymm2 \n"
"vpsrlw $0x7,%%ymm0,%%ymm0 \n"
"vpsrlw $0x7,%%ymm2,%%ymm2 \n"
"vpackuswb %%ymm2,%%ymm0,%%ymm0 \n"
"vpermd %%ymm0,%%ymm6,%%ymm0 \n"
"vpaddb %%ymm5,%%ymm0,%%ymm0 \n"
"vpackuswb %%ymm2,%%ymm0,%%ymm0 \n" // mutates.
"vpermd %%ymm0,%%ymm6,%%ymm0 \n" // unmutate.
"vpaddb %%ymm5,%%ymm0,%%ymm0 \n" // add 16 for Y
"vmovdqu %%ymm0," MEMACCESS(1) " \n"
"lea " MEMLEA(0x20,1) ",%1 \n"
"sub $0x20,%2 \n"
......@@ -831,6 +836,7 @@ void ARGBToYRow_AVX2(const uint8* src_argb, uint8* dst_y, int pix) {
#endif // HAS_ARGBTOYROW_AVX2
#ifdef HAS_ARGBTOYJROW_AVX2
// Convert 32 ARGB pixels (128 bytes) to 32 Y values.
void ARGBToYJRow_AVX2(const uint8* src_argb, uint8* dst_y, int pix) {
asm volatile (
"vbroadcastf128 %3,%%ymm4 \n"
......@@ -847,13 +853,14 @@ void ARGBToYJRow_AVX2(const uint8* src_argb, uint8* dst_y, int pix) {
"vpmaddubsw %%ymm4,%%ymm2,%%ymm2 \n"
"vpmaddubsw %%ymm4,%%ymm3,%%ymm3 \n"
"lea " MEMLEA(0x80,0) ",%0 \n"
"vphaddw %%ymm1,%%ymm0,%%ymm0 \n"
"vphaddw %%ymm1,%%ymm0,%%ymm0 \n" // mutates.
"vphaddw %%ymm3,%%ymm2,%%ymm2 \n"
"vpaddw %%ymm5,%%ymm0,%%ymm0 \n" // Add .5 for rounding.
"vpaddw %%ymm5,%%ymm2,%%ymm2 \n"
"vpsrlw $0x7,%%ymm0,%%ymm0 \n"
"vpsrlw $0x7,%%ymm2,%%ymm2 \n"
"vpackuswb %%ymm2,%%ymm0,%%ymm0 \n"
"vpermd %%ymm0,%%ymm6,%%ymm0 \n"
"vpaddb %%ymm5,%%ymm0,%%ymm0 \n"
"vpackuswb %%ymm2,%%ymm0,%%ymm0 \n" // mutates.
"vpermd %%ymm0,%%ymm6,%%ymm0 \n" // unmutate.
"vmovdqu %%ymm0," MEMACCESS(1) " \n"
"lea " MEMLEA(0x20,1) ",%1 \n"
"sub $0x20,%2 \n"
......@@ -863,7 +870,7 @@ void ARGBToYJRow_AVX2(const uint8* src_argb, uint8* dst_y, int pix) {
"+r"(dst_y), // %1
"+r"(pix) // %2
: "m"(kARGBToYJ), // %3
"m"(kAddY16), // %4
"m"(kAddYJ64), // %4
"m"(kPermdARGBToY_AVX) // %5
: "memory", "cc"
#if defined(__SSE2__)
......
source/row_win.cc
View file @ 540e8af8
......@@ -210,6 +210,7 @@ static const uvec8 kAddY16 = {
16u, 16u, 16u, 16u, 16u, 16u, 16u, 16u, 16u, 16u, 16u, 16u, 16u, 16u, 16u, 16u
};
// 7 bit fixed point 0.5.
static const vec16 kAddYJ64 = {
64, 64, 64, 64, 64, 64, 64, 64
};
......@@ -697,8 +698,8 @@ void ARGBToYRow_SSSE3(const uint8* src_argb, uint8* dst_y, int pix) {
mov eax, [esp + 4] /* src_argb */
mov edx, [esp + 8] /* dst_y */
mov ecx, [esp + 12] /* pix */
movdqa xmm5, kAddY16
movdqa xmm4, kARGBToY
movdqa xmm5, kAddY16
convertloop:
movdqu xmm0, [eax]
......@@ -724,7 +725,8 @@ void ARGBToYRow_SSSE3(const uint8* src_argb, uint8* dst_y, int pix) {
}
}
// Convert 16 ARGB pixels (64 bytes) to 16 Y values.
// Convert 16 ARGB pixels (64 bytes) to 16 YJ values.
// Same as ARGBToYRow but different coefficients, no add 16, but do rounding.
__declspec(naked) __declspec(align(16))
void ARGBToYJRow_SSSE3(const uint8* src_argb, uint8* dst_y, int pix) {
__asm {
......@@ -787,7 +789,7 @@ void ARGBToYRow_AVX2(const uint8* src_argb, uint8* dst_y, int pix) {
vpsrlw ymm2, ymm2, 7
vpackuswb ymm0, ymm0, ymm2 // mutates.
vpermd ymm0, ymm6, ymm0 // For vphaddw + vpackuswb mutation.
vpaddb ymm0, ymm0, ymm5
vpaddb ymm0, ymm0, ymm5 // add 16 for Y
vmovdqu [edx], ymm0
lea edx, [edx + 32]
sub ecx, 32
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment