Skip to content

L
libyuv
Commit 714050a2 authored by fbarchard@google.com's avatar fbarchard@google.com
Browse files
Options

sse version of BGRA and ABGR To I420 

BUG=none
TEST=none
Review URL: https://webrtc-codereview.appspot.com/400004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@178 16f28f9a-4ce2-e073-06de-1de4eb20be90
parent 69fe6bd1
Show whitespace changes
Inline Side-by-side
Showing with 471 additions and 81 deletions
DEPS
View file @ 714050a2
vars = {
"libyuv_trunk" : "https://libyuv.googlecode.com/svn/trunk",
"chromium_trunk" : "http://src.chromium.org/svn/trunk",
"chromium_revision": "119959",
"chromium_revision": "95033",
# Use this googlecode_url variable only if there is an internal mirror for it.
# If you do not know, use the full path while defining your new deps entry.
"googlecode_url": "http://%s.googlecode.com/svn",
......@@ -22,13 +22,12 @@ deps = {
# Dependencies used by libjpeg-turbo
"trunk/third_party/libjpeg_turbo/":
Var("chromium_trunk") + "/src/third_party/libjpeg_turbo@" + Var("chromium_revision"),
Var("chromium_trunk") + "/deps/third_party/libjpeg_turbo@119959",
"trunk/third_party/yasm/":
Var("chromium_trunk") + "/src/third_party/yasm@" + Var("chromium_revision"),
}
hooks = [
# A change to a .gyp, .gypi, or to GYP itself should run the generator.
{
......
README.chromium
View file @ 714050a2
Name: libyuv
URL: http://code.google.com/p/libyuv/
Version: 177
Version: 178
License: BSD
License File: LICENSE
......
include/libyuv/convert.h
View file @ 714050a2
......@@ -156,6 +156,17 @@ int ARGB4444ToI420(const uint8* src_frame, int src_stride_frame,
uint8* dst_v, int dst_stride_v,
int width, int height);
#ifdef HAVE_JPEG
// src_width/height provided by capture
// dst_width/height for clipping determine final size.
int MJPGToI420(const uint8* sample, size_t sample_size,
uint8* dst_y, int dst_stride_y,
uint8* dst_u, int dst_stride_u,
uint8* dst_v, int dst_stride_v,
int src_width, int src_height,
int dst_width, int dst_height);
#endif
// Note Bayer formats (BGGR) To I420 are in format_conversion.h
// Convert camera sample to I420 with cropping, rotation and vertical flip.
......
include/libyuv/version.h
View file @ 714050a2
......@@ -11,7 +11,7 @@
#ifndef INCLUDE_LIBYUV_VERSION_H_
#define INCLUDE_LIBYUV_VERSION_H_
#define LIBYUV_VERSION 177
#define LIBYUV_VERSION 178
#endif // INCLUDE_LIBYUV_VERSION_H_
source/row_posix.cc
View file @ 714050a2
......@@ -20,38 +20,64 @@ extern "C" {
// This module is for GCC x86 and x64
#if (defined(__x86_64__) || defined(__i386__)) && !defined(YUV_DISABLE_ASM)
// GCC 4.2 on OSX has link error when passing static or const to inline.
// TODO(fbarchard): Use static const when gcc 4.2 support is dropped.
#ifdef __APPLE__
#define CONST
#else
#define CONST static const
#endif
#ifdef HAS_ARGBTOUVROW_SSSE3
#ifdef HAS_ARGBTOYROW_SSSE3
// Constants for ARGB
CONST vec8 kARGBToY = {
13, 65, 33, 0, 13, 65, 33, 0, 13, 65, 33, 0, 13, 65, 33, 0
};
CONST vec8 kARGBToU = {
112, -74, -38, 0, 112, -74, -38, 0, 112, -74, -38, 0, 112, -74, -38, 0
};
CONST uvec8 kARGBToV = {
-18, -94, 112, 0, -18, -94, 112, 0, -18, -94, 112, 0, -18, -94, 112, 0
CONST vec8 kARGBToV = {
-18, -94, 112, 0, -18, -94, 112, 0, -18, -94, 112, 0, -18, -94, 112, 0,
};
CONST uvec8 kAddUV128 = {
128u, 128u, 128u, 128u, 128u, 128u, 128u, 128u,
128u, 128u, 128u, 128u, 128u, 128u, 128u, 128u
// Constants for BGRA
CONST vec8 kBGRAToY = {
0, 33, 65, 13, 0, 33, 65, 13, 0, 33, 65, 13, 0, 33, 65, 13
};
#endif
#ifdef HAS_ARGBTOYROW_SSSE3
CONST vec8 kBGRAToU = {
0, -38, -74, 112, 0, -38, -74, 112, 0, -38, -74, 112, 0, -38, -74, 112
};
// Constant multiplication table for converting ARGB to I400.
CONST vec8 kARGBToY = {
13, 65, 33, 0, 13, 65, 33, 0, 13, 65, 33, 0, 13, 65, 33, 0
CONST vec8 kBGRAToV = {
0, 112, -94, -18, 0, 112, -94, -18, 0, 112, -94, -18, 0, 112, -94, -18
};
// Constants for ABGR
CONST vec8 kABGRToY = {
33, 65, 13, 0, 33, 65, 13, 0, 33, 65, 13, 0, 33, 65, 13, 0
};
CONST vec8 kABGRToU = {
-38, -74, 112, 0, -38, -74, 112, 0, -38, -74, 112, 0, -38, -74, 112, 0
};
CONST vec8 kABGRToV = {
112, -94, -18, 0, 112, -94, -18, 0, 112, -94, -18, 0, 112, -94, -18, 0
};
CONST uvec8 kAddY16 = {
16u, 16u, 16u, 16u, 16u, 16u, 16u, 16u, 16u, 16u, 16u, 16u, 16u, 16u, 16u, 16u
};
CONST uvec8 kAddUV128 = {
128u, 128u, 128u, 128u, 128u, 128u, 128u, 128u,
128u, 128u, 128u, 128u, 128u, 128u, 128u, 128u
};
// Shuffle table for converting RGB24 to ARGB.
CONST uvec8 kShuffleMaskRGB24ToARGB = {
0u, 1u, 2u, 12u, 3u, 4u, 5u, 13u, 6u, 7u, 8u, 14u, 9u, 10u, 11u, 15u
......@@ -77,7 +103,6 @@ CONST uvec8 kShuffleMaskARGBToRGB24 = {
0u, 1u, 2u, 4u, 5u, 6u, 8u, 9u, 10u, 12u, 13u, 14u, 128u, 128u, 128u, 128u
};
// Shuffle table for converting ARGB to RAW.
CONST uvec8 kShuffleMaskARGBToRAW = {
2u, 1u,0u, 6u, 5u, 4u, 10u, 9u, 8u, 14u, 13u, 12u, 128u, 128u, 128u, 128u
......@@ -569,7 +594,6 @@ void ARGBToARGB4444Row_SSE2(const uint8* src, uint8* dst, int pix) {
);
}
void ARGBToYRow_SSSE3(const uint8* src_argb, uint8* dst_y, int pix) {
asm volatile (
"movdqa %4,%%xmm5 \n"
......@@ -641,9 +665,12 @@ void ARGBToYRow_Unaligned_SSSE3(const uint8* src_argb, uint8* dst_y, int pix) {
#endif
);
}
#endif
#ifdef HAS_ARGBTOUVROW_SSSE3
// TODO(fbarchard): pass xmm constants to single block of assembly.
// fpic on GCC 4.2 for OSX runs out of GPR registers. "m" effectively takes
// 3 registers - ebx, ebp and eax. "m" can be passed with 3 normal registers,
// or 4 if stack frame is disabled. Doing 2 assembly blocks is a work around
// and considered unsafe.
void ARGBToUVRow_SSSE3(const uint8* src_argb0, int src_stride_argb,
uint8* dst_u, uint8* dst_v, int width) {
asm volatile (
......@@ -775,7 +802,420 @@ void ARGBToUVRow_Unaligned_SSSE3(const uint8* src_argb0, int src_stride_argb,
#endif
);
}
void BGRAToYRow_SSSE3(const uint8* src_bgra, uint8* dst_y, int pix) {
asm volatile (
"movdqa %4,%%xmm5 \n"
"movdqa %3,%%xmm4 \n"
"1: \n"
"movdqa (%0),%%xmm0 \n"
"movdqa 0x10(%0),%%xmm1 \n"
"movdqa 0x20(%0),%%xmm2 \n"
"movdqa 0x30(%0),%%xmm3 \n"
"pmaddubsw %%xmm4,%%xmm0 \n"
"pmaddubsw %%xmm4,%%xmm1 \n"
"pmaddubsw %%xmm4,%%xmm2 \n"
"pmaddubsw %%xmm4,%%xmm3 \n"
"lea 0x40(%0),%0 \n"
"phaddw %%xmm1,%%xmm0 \n"
"phaddw %%xmm3,%%xmm2 \n"
"psrlw $0x7,%%xmm0 \n"
"psrlw $0x7,%%xmm2 \n"
"packuswb %%xmm2,%%xmm0 \n"
"paddb %%xmm5,%%xmm0 \n"
"movdqa %%xmm0,(%1) \n"
"lea 0x10(%1),%1 \n"
"sub $0x10,%2 \n"
"ja 1b \n"
: "+r"(src_bgra), // %0
"+r"(dst_y), // %1
"+r"(pix) // %2
: "m"(kBGRAToY), // %3
"m"(kAddY16) // %4
: "memory", "cc"
#if defined(__SSE2__)
, "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5"
#endif
);
}
void BGRAToYRow_Unaligned_SSSE3(const uint8* src_bgra, uint8* dst_y, int pix) {
asm volatile (
"movdqa %4,%%xmm5 \n"
"movdqa %3,%%xmm4 \n"
"1: \n"
"movdqu (%0),%%xmm0 \n"
"movdqu 0x10(%0),%%xmm1 \n"
"movdqu 0x20(%0),%%xmm2 \n"
"movdqu 0x30(%0),%%xmm3 \n"
"pmaddubsw %%xmm4,%%xmm0 \n"
"pmaddubsw %%xmm4,%%xmm1 \n"
"pmaddubsw %%xmm4,%%xmm2 \n"
"pmaddubsw %%xmm4,%%xmm3 \n"
"lea 0x40(%0),%0 \n"
"phaddw %%xmm1,%%xmm0 \n"
"phaddw %%xmm3,%%xmm2 \n"
"psrlw $0x7,%%xmm0 \n"
"psrlw $0x7,%%xmm2 \n"
"packuswb %%xmm2,%%xmm0 \n"
"paddb %%xmm5,%%xmm0 \n"
"movdqu %%xmm0,(%1) \n"
"lea 0x10(%1),%1 \n"
"sub $0x10,%2 \n"
"ja 1b \n"
: "+r"(src_bgra), // %0
"+r"(dst_y), // %1
"+r"(pix) // %2
: "m"(kBGRAToY), // %3
"m"(kAddY16) // %4
: "memory", "cc"
#if defined(__SSE2__)
, "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5"
#endif
);
}
void BGRAToUVRow_SSSE3(const uint8* src_bgra0, int src_stride_bgra,
uint8* dst_u, uint8* dst_v, int width) {
asm volatile (
"movdqa %0,%%xmm4 \n"
"movdqa %1,%%xmm3 \n"
"movdqa %2,%%xmm5 \n"
:
: "m"(kBGRAToU), // %0
"m"(kBGRAToV), // %1
"m"(kAddUV128) // %2
:
#if defined(__SSE2__)
"xmm3", "xmm4", "xmm5"
#endif
);
asm volatile (
"sub %1,%2 \n"
"1: \n"
"movdqa (%0),%%xmm0 \n"
"movdqa 0x10(%0),%%xmm1 \n"
"movdqa 0x20(%0),%%xmm2 \n"
"movdqa 0x30(%0),%%xmm6 \n"
"pavgb (%0,%4,1),%%xmm0 \n"
"pavgb 0x10(%0,%4,1),%%xmm1 \n"
"pavgb 0x20(%0,%4,1),%%xmm2 \n"
"pavgb 0x30(%0,%4,1),%%xmm6 \n"
"lea 0x40(%0),%0 \n"
"movdqa %%xmm0,%%xmm7 \n"
"shufps $0x88,%%xmm1,%%xmm0 \n"
"shufps $0xdd,%%xmm1,%%xmm7 \n"
"pavgb %%xmm7,%%xmm0 \n"
"movdqa %%xmm2,%%xmm7 \n"
"shufps $0x88,%%xmm6,%%xmm2 \n"
"shufps $0xdd,%%xmm6,%%xmm7 \n"
"pavgb %%xmm7,%%xmm2 \n"
"movdqa %%xmm0,%%xmm1 \n"
"movdqa %%xmm2,%%xmm6 \n"
"pmaddubsw %%xmm4,%%xmm0 \n"
"pmaddubsw %%xmm4,%%xmm2 \n"
"pmaddubsw %%xmm3,%%xmm1 \n"
"pmaddubsw %%xmm3,%%xmm6 \n"
"phaddw %%xmm2,%%xmm0 \n"
"phaddw %%xmm6,%%xmm1 \n"
"psraw $0x8,%%xmm0 \n"
"psraw $0x8,%%xmm1 \n"
"packsswb %%xmm1,%%xmm0 \n"
"paddb %%xmm5,%%xmm0 \n"
"movlps %%xmm0,(%1) \n"
"movhps %%xmm0,(%1,%2,1) \n"
"lea 0x8(%1),%1 \n"
"sub $0x10,%3 \n"
"ja 1b \n"
: "+r"(src_bgra0), // %0
"+r"(dst_u), // %1
"+r"(dst_v), // %2
"+rm"(width) // %3
: "r"(static_cast<intptr_t>(src_stride_bgra))
: "memory", "cc"
#if defined(__SSE2__)
, "xmm0", "xmm1", "xmm2", "xmm6", "xmm7"
#endif
);
}
void BGRAToUVRow_Unaligned_SSSE3(const uint8* src_bgra0, int src_stride_bgra,
uint8* dst_u, uint8* dst_v, int width) {
asm volatile (
"movdqa %0,%%xmm4 \n"
"movdqa %1,%%xmm3 \n"
"movdqa %2,%%xmm5 \n"
:
: "m"(kBGRAToU), // %0
"m"(kBGRAToV), // %1
"m"(kAddUV128) // %2
:
#if defined(__SSE2__)
"xmm3", "xmm4", "xmm5"
#endif
);
asm volatile (
"sub %1,%2 \n"
"1: \n"
"movdqu (%0),%%xmm0 \n"
"movdqu 0x10(%0),%%xmm1 \n"
"movdqu 0x20(%0),%%xmm2 \n"
"movdqu 0x30(%0),%%xmm6 \n"
"movdqu (%0,%4,1),%%xmm7 \n"
"pavgb %%xmm7,%%xmm0 \n"
"movdqu 0x10(%0,%4,1),%%xmm7 \n"
"pavgb %%xmm7,%%xmm1 \n"
"movdqu 0x20(%0,%4,1),%%xmm7 \n"
"pavgb %%xmm7,%%xmm2 \n"
"movdqu 0x30(%0,%4,1),%%xmm7 \n"
"pavgb %%xmm7,%%xmm6 \n"
"lea 0x40(%0),%0 \n"
"movdqa %%xmm0,%%xmm7 \n"
"shufps $0x88,%%xmm1,%%xmm0 \n"
"shufps $0xdd,%%xmm1,%%xmm7 \n"
"pavgb %%xmm7,%%xmm0 \n"
"movdqa %%xmm2,%%xmm7 \n"
"shufps $0x88,%%xmm6,%%xmm2 \n"
"shufps $0xdd,%%xmm6,%%xmm7 \n"
"pavgb %%xmm7,%%xmm2 \n"
"movdqa %%xmm0,%%xmm1 \n"
"movdqa %%xmm2,%%xmm6 \n"
"pmaddubsw %%xmm4,%%xmm0 \n"
"pmaddubsw %%xmm4,%%xmm2 \n"
"pmaddubsw %%xmm3,%%xmm1 \n"
"pmaddubsw %%xmm3,%%xmm6 \n"
"phaddw %%xmm2,%%xmm0 \n"
"phaddw %%xmm6,%%xmm1 \n"
"psraw $0x8,%%xmm0 \n"
"psraw $0x8,%%xmm1 \n"
"packsswb %%xmm1,%%xmm0 \n"
"paddb %%xmm5,%%xmm0 \n"
"movlps %%xmm0,(%1) \n"
"movhps %%xmm0,(%1,%2,1) \n"
"lea 0x8(%1),%1 \n"
"sub $0x10,%3 \n"
"ja 1b \n"
: "+r"(src_bgra0), // %0
"+r"(dst_u), // %1
"+r"(dst_v), // %2
"+rm"(width) // %3
: "r"(static_cast<intptr_t>(src_stride_bgra))
: "memory", "cc"
#if defined(__SSE2__)
, "xmm0", "xmm1", "xmm2", "xmm6", "xmm7"
#endif
);
}
#endif
void ABGRToYRow_SSSE3(const uint8* src_abgr, uint8* dst_y, int pix) {
asm volatile (
"movdqa %4,%%xmm5 \n"
"movdqa %3,%%xmm4 \n"
"1: \n"
"movdqa (%0),%%xmm0 \n"
"movdqa 0x10(%0),%%xmm1 \n"
"movdqa 0x20(%0),%%xmm2 \n"
"movdqa 0x30(%0),%%xmm3 \n"
"pmaddubsw %%xmm4,%%xmm0 \n"
"pmaddubsw %%xmm4,%%xmm1 \n"
"pmaddubsw %%xmm4,%%xmm2 \n"
"pmaddubsw %%xmm4,%%xmm3 \n"
"lea 0x40(%0),%0 \n"
"phaddw %%xmm1,%%xmm0 \n"
"phaddw %%xmm3,%%xmm2 \n"
"psrlw $0x7,%%xmm0 \n"
"psrlw $0x7,%%xmm2 \n"
"packuswb %%xmm2,%%xmm0 \n"
"paddb %%xmm5,%%xmm0 \n"
"movdqa %%xmm0,(%1) \n"
"lea 0x10(%1),%1 \n"
"sub $0x10,%2 \n"
"ja 1b \n"
: "+r"(src_abgr), // %0
"+r"(dst_y), // %1
"+r"(pix) // %2
: "m"(kABGRToY), // %3
"m"(kAddY16) // %4
: "memory", "cc"
#if defined(__SSE2__)
, "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5"
#endif
);
}
void ABGRToYRow_Unaligned_SSSE3(const uint8* src_abgr, uint8* dst_y, int pix) {
asm volatile (
"movdqa %4,%%xmm5 \n"
"movdqa %3,%%xmm4 \n"
"1: \n"
"movdqu (%0),%%xmm0 \n"
"movdqu 0x10(%0),%%xmm1 \n"
"movdqu 0x20(%0),%%xmm2 \n"
"movdqu 0x30(%0),%%xmm3 \n"
"pmaddubsw %%xmm4,%%xmm0 \n"
"pmaddubsw %%xmm4,%%xmm1 \n"
"pmaddubsw %%xmm4,%%xmm2 \n"
"pmaddubsw %%xmm4,%%xmm3 \n"
"lea 0x40(%0),%0 \n"
"phaddw %%xmm1,%%xmm0 \n"
"phaddw %%xmm3,%%xmm2 \n"
"psrlw $0x7,%%xmm0 \n"
"psrlw $0x7,%%xmm2 \n"
"packuswb %%xmm2,%%xmm0 \n"
"paddb %%xmm5,%%xmm0 \n"
"movdqu %%xmm0,(%1) \n"
"lea 0x10(%1),%1 \n"
"sub $0x10,%2 \n"
"ja 1b \n"
: "+r"(src_abgr), // %0
"+r"(dst_y), // %1
"+r"(pix) // %2
: "m"(kABGRToY), // %3
"m"(kAddY16) // %4
: "memory", "cc"
#if defined(__SSE2__)
, "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5"
#endif
);
}
void ABGRToUVRow_SSSE3(const uint8* src_abgr0, int src_stride_abgr,
uint8* dst_u, uint8* dst_v, int width) {
asm volatile (
"movdqa %0,%%xmm4 \n"
"movdqa %1,%%xmm3 \n"
"movdqa %2,%%xmm5 \n"
:
: "m"(kABGRToU), // %0
"m"(kABGRToV), // %1
"m"(kAddUV128) // %2
:
#if defined(__SSE2__)
"xmm3", "xmm4", "xmm5"
#endif
);
asm volatile (
"sub %1,%2 \n"
"1: \n"
"movdqa (%0),%%xmm0 \n"
"movdqa 0x10(%0),%%xmm1 \n"
"movdqa 0x20(%0),%%xmm2 \n"
"movdqa 0x30(%0),%%xmm6 \n"
"pavgb (%0,%4,1),%%xmm0 \n"
"pavgb 0x10(%0,%4,1),%%xmm1 \n"
"pavgb 0x20(%0,%4,1),%%xmm2 \n"
"pavgb 0x30(%0,%4,1),%%xmm6 \n"
"lea 0x40(%0),%0 \n"
"movdqa %%xmm0,%%xmm7 \n"
"shufps $0x88,%%xmm1,%%xmm0 \n"
"shufps $0xdd,%%xmm1,%%xmm7 \n"
"pavgb %%xmm7,%%xmm0 \n"
"movdqa %%xmm2,%%xmm7 \n"
"shufps $0x88,%%xmm6,%%xmm2 \n"
"shufps $0xdd,%%xmm6,%%xmm7 \n"
"pavgb %%xmm7,%%xmm2 \n"
"movdqa %%xmm0,%%xmm1 \n"
"movdqa %%xmm2,%%xmm6 \n"
"pmaddubsw %%xmm4,%%xmm0 \n"
"pmaddubsw %%xmm4,%%xmm2 \n"
"pmaddubsw %%xmm3,%%xmm1 \n"
"pmaddubsw %%xmm3,%%xmm6 \n"
"phaddw %%xmm2,%%xmm0 \n"
"phaddw %%xmm6,%%xmm1 \n"
"psraw $0x8,%%xmm0 \n"
"psraw $0x8,%%xmm1 \n"
"packsswb %%xmm1,%%xmm0 \n"
"paddb %%xmm5,%%xmm0 \n"
"movlps %%xmm0,(%1) \n"
"movhps %%xmm0,(%1,%2,1) \n"
"lea 0x8(%1),%1 \n"
"sub $0x10,%3 \n"
"ja 1b \n"
: "+r"(src_abgr0), // %0
"+r"(dst_u), // %1
"+r"(dst_v), // %2
"+rm"(width) // %3
: "r"(static_cast<intptr_t>(src_stride_abgr))
: "memory", "cc"
#if defined(__SSE2__)
, "xmm0", "xmm1", "xmm2", "xmm6", "xmm7"
#endif
);
}
void ABGRToUVRow_Unaligned_SSSE3(const uint8* src_abgr0, int src_stride_abgr,
uint8* dst_u, uint8* dst_v, int width) {
asm volatile (
"movdqa %0,%%xmm4 \n"
"movdqa %1,%%xmm3 \n"
"movdqa %2,%%xmm5 \n"
:
: "m"(kABGRToU), // %0
"m"(kABGRToV), // %1
"m"(kAddUV128) // %2
:
#if defined(__SSE2__)
"xmm3", "xmm4", "xmm5"
#endif
);
asm volatile (
"sub %1,%2 \n"
"1: \n"
"movdqu (%0),%%xmm0 \n"
"movdqu 0x10(%0),%%xmm1 \n"
"movdqu 0x20(%0),%%xmm2 \n"
"movdqu 0x30(%0),%%xmm6 \n"
"movdqu (%0,%4,1),%%xmm7 \n"
"pavgb %%xmm7,%%xmm0 \n"
"movdqu 0x10(%0,%4,1),%%xmm7 \n"
"pavgb %%xmm7,%%xmm1 \n"
"movdqu 0x20(%0,%4,1),%%xmm7 \n"
"pavgb %%xmm7,%%xmm2 \n"
"movdqu 0x30(%0,%4,1),%%xmm7 \n"
"pavgb %%xmm7,%%xmm6 \n"
"lea 0x40(%0),%0 \n"
"movdqa %%xmm0,%%xmm7 \n"
"shufps $0x88,%%xmm1,%%xmm0 \n"
"shufps $0xdd,%%xmm1,%%xmm7 \n"
"pavgb %%xmm7,%%xmm0 \n"
"movdqa %%xmm2,%%xmm7 \n"
"shufps $0x88,%%xmm6,%%xmm2 \n"
"shufps $0xdd,%%xmm6,%%xmm7 \n"
"pavgb %%xmm7,%%xmm2 \n"
"movdqa %%xmm0,%%xmm1 \n"
"movdqa %%xmm2,%%xmm6 \n"
"pmaddubsw %%xmm4,%%xmm0 \n"
"pmaddubsw %%xmm4,%%xmm2 \n"
"pmaddubsw %%xmm3,%%xmm1 \n"
"pmaddubsw %%xmm3,%%xmm6 \n"
"phaddw %%xmm2,%%xmm0 \n"
"phaddw %%xmm6,%%xmm1 \n"
"psraw $0x8,%%xmm0 \n"
"psraw $0x8,%%xmm1 \n"
"packsswb %%xmm1,%%xmm0 \n"
"paddb %%xmm5,%%xmm0 \n"
"movlps %%xmm0,(%1) \n"
"movhps %%xmm0,(%1,%2,1) \n"
"lea 0x8(%1),%1 \n"
"sub $0x10,%3 \n"
"ja 1b \n"
: "+r"(src_abgr0), // %0
"+r"(dst_u), // %1
"+r"(dst_v), // %2
"+rm"(width) // %3
: "r"(static_cast<intptr_t>(src_stride_abgr))
: "memory", "cc"
#if defined(__SSE2__)
, "xmm0", "xmm1", "xmm2", "xmm6", "xmm7"
#endif
);
}
#endif
#endif // HAS_ARGBTOYROW_SSSE3
#ifdef HAS_I420TOARGBROW_SSSE3
#define UB 127 /* min(63,static_cast<int8>(2.018 * 64)) */
......@@ -1056,66 +1496,6 @@ void YToARGBRow_SSE2(const uint8* y_buf,
}
#endif
#ifdef HAS_ARGBTOYROW_SSSE3
void ABGRToYRow_SSSE3(const uint8* src_argb, uint8* dst_y, int pix) {
SIMD_ALIGNED(uint8 row[kMaxStride]);
ABGRToARGBRow_SSSE3(src_argb, row, pix);
ARGBToYRow_SSSE3(row, dst_y, pix);
}
void BGRAToYRow_SSSE3(const uint8* src_argb, uint8* dst_y, int pix) {
SIMD_ALIGNED(uint8 row[kMaxStride]);
BGRAToARGBRow_SSSE3(src_argb, row, pix);
ARGBToYRow_SSSE3(row, dst_y, pix);
}
void ABGRToYRow_Unaligned_SSSE3(const uint8* src_argb, uint8* dst_y, int pix) {
SIMD_ALIGNED(uint8 row[kMaxStride]);
ABGRToARGBRow_C(src_argb, row, pix);
ARGBToYRow_SSSE3(row, dst_y, pix);
}
void BGRAToYRow_Unaligned_SSSE3(const uint8* src_argb, uint8* dst_y, int pix) {
SIMD_ALIGNED(uint8 row[kMaxStride]);
BGRAToARGBRow_C(src_argb, row, pix);
ARGBToYRow_SSSE3(row, dst_y, pix);
}
#endif
#ifdef HAS_ARGBTOUVROW_SSSE3
void ABGRToUVRow_SSSE3(const uint8* src_argb, int src_stride_argb,
uint8* dst_u, uint8* dst_v, int pix) {
SIMD_ALIGNED(uint8 row[kMaxStride * 2]);
ABGRToARGBRow_SSSE3(src_argb, row, pix);
ABGRToARGBRow_SSSE3(src_argb + src_stride_argb, row + kMaxStride, pix);
ARGBToUVRow_SSSE3(row, kMaxStride, dst_u, dst_v, pix);
}
void BGRAToUVRow_SSSE3(const uint8* src_argb, int src_stride_argb,
uint8* dst_u, uint8* dst_v, int pix) {
SIMD_ALIGNED(uint8 row[kMaxStride * 2]);
BGRAToARGBRow_SSSE3(src_argb, row, pix);
BGRAToARGBRow_SSSE3(src_argb + src_stride_argb, row + kMaxStride, pix);
ARGBToUVRow_SSSE3(row, kMaxStride, dst_u, dst_v, pix);
}
void ABGRToUVRow_Unaligned_SSSE3(const uint8* src_argb, int src_stride_argb,
uint8* dst_u, uint8* dst_v, int pix) {
SIMD_ALIGNED(uint8 row[kMaxStride * 2]);
ABGRToARGBRow_C(src_argb, row, pix);
ABGRToARGBRow_C(src_argb + src_stride_argb, row + kMaxStride, pix);
ARGBToUVRow_SSSE3(row, kMaxStride, dst_u, dst_v, pix);
}
void BGRAToUVRow_Unaligned_SSSE3(const uint8* src_argb, int src_stride_argb,
uint8* dst_u, uint8* dst_v, int pix) {
SIMD_ALIGNED(uint8 row[kMaxStride * 2]);
BGRAToARGBRow_C(src_argb, row, pix);
BGRAToARGBRow_C(src_argb + src_stride_argb, row + kMaxStride, pix);
ARGBToUVRow_SSSE3(row, kMaxStride, dst_u, dst_v, pix);
}
#endif
#ifdef HAS_MIRRORROW_SSSE3
// Shuffle table for reversing the bytes.
......
source/row_win.cc
View file @ 714050a2
......@@ -20,7 +20,7 @@ extern "C" {
#ifdef HAS_ARGBTOYROW_SSSE3
// Constant multiplication table for converting ARGB to I400.
// Constants for ARGB
static const vec8 kARGBToY = {
13, 65, 33, 0, 13, 65, 33, 0, 13, 65, 33, 0, 13, 65, 33, 0
};
......
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