coffee/libyuv
1
0
Fork 0
mirror of https://chromium.googlesource.com/libyuv/libyuv synced 2025-12-07 01:06:46 +08:00
Code Issues Packages Projects Releases Wiki Activity

Port Even ARGBScale to gcc

Browse Source 
BUG=none
TEST=none
Review URL: https://webrtc-codereview.appspot.com/566010

git-svn-id: http://libyuv.googlecode.com/svn/trunk@264 16f28f9a-4ce2-e073-06de-1de4eb20be90
This commit is contained in:
fbarchard@google.com 2012-05-09 22:52:23 +00:00
parent c7bd1dbea6
commit 68caaed871
3 changed files with 88 additions and 2 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
URL: http://code.google.com/p/libyuv/
Version: 263
Version: 264
License: BSD
License File: LICENSE

2
include/libyuv/version.h
View File

@ -11,7 +11,7 @@
#ifndef INCLUDE_LIBYUV_VERSION_H_
#define INCLUDE_LIBYUV_VERSION_H_
#define LIBYUV_VERSION 263
#define LIBYUV_VERSION 264
#endif // INCLUDE_LIBYUV_VERSION_H_

86
source/scale_argb.cc
View File

@ -411,6 +411,92 @@ static void ScaleARGBRowDown2Int_SSE2(const uint8* src_ptr, int src_stride,
);
}
#define HAS_SCALEARGBROWDOWNEVEN_SSE2
// Reads 4 pixels at a time.
// Alignment requirement: dst_ptr 16 byte aligned.
static void ScaleARGBRowDownEven_SSE2(const uint8* src_ptr, int src_stride,
int src_stepx,
uint8* dst_ptr, int dst_width) {
intptr_t src_stepx_x4 = static_cast<intptr_t>(src_stepx);
intptr_t src_stepx_x12 = 0;
asm volatile (
"lea 0x0(,%1,4),%1 \n"
"lea (%1,%1,2),%4 \n"
".p2align 4 \n"
"1: \n"
"movd (%0),%%xmm0 \n"
"movd (%0,%1,1),%%xmm1 \n"
"punpckldq %%xmm1,%%xmm0 \n"
"movd (%0,%1,2),%%xmm2 \n"
"movd (%0,%4,1),%%xmm3 \n"
"lea (%0,%1,4),%0 \n"
"punpckldq %%xmm3,%%xmm2 \n"
"punpcklqdq %%xmm2,%%xmm0 \n"
"sub $0x4,%3 \n"
"movdqa %%xmm0,(%2) \n"
"lea 0x10(%2),%2 \n"
"jg 1b \n"
: "+r"(src_ptr), // %0
"+r"(src_stepx_x4), // %1
"+r"(dst_ptr), // %2
"+r"(dst_width), // %3
"+r"(src_stepx_x12) // %4
:
: "memory", "cc"
#if defined(__SSE2__)
, "xmm0", "xmm1", "xmm2", "xmm3"
#endif
);
}
// Blends four 2x2 to 4x1.
// Alignment requirement: dst_ptr 16 byte aligned.
static void ScaleARGBRowDownEvenInt_SSE2(const uint8* src_ptr, int src_stride,
int src_stepx,
uint8* dst_ptr, int dst_width) {
intptr_t src_stepx_x4 = static_cast<intptr_t>(src_stepx);
intptr_t src_stepx_x12 = 0;
intptr_t row1 = static_cast<intptr_t>(src_stride);
asm volatile (
"lea 0x0(,%1,4),%1 \n"
"lea (%1,%1,2),%4 \n"
"lea (%0,%5,1),%5 \n"
".p2align 4 \n"
"1: \n"
"movq (%0),%%xmm0 \n"
"movhps (%0,%1,1),%%xmm0 \n"
"movq (%0,%1,2),%%xmm1 \n"
"movhps (%0,%4,1),%%xmm1 \n"
"lea (%0,%1,4),%0 \n"
"movq (%5),%%xmm2 \n"
"movhps (%5,%1,1),%%xmm2 \n"
"movq (%5,%1,2),%%xmm3 \n"
"movhps (%5,%4,1),%%xmm3 \n"
"lea (%5,%1,4),%5 \n"
"pavgb %%xmm2,%%xmm0 \n"
"pavgb %%xmm3,%%xmm1 \n"
"movdqa %%xmm0,%%xmm2 \n"
"shufps $0x88,%%xmm1,%%xmm0 \n"
"shufps $0xdd,%%xmm1,%%xmm2 \n"
"pavgb %%xmm2,%%xmm0 \n"
"sub $0x4,%3 \n"
"movdqa %%xmm0,(%2) \n"
"lea 0x10(%2),%2 \n"
"jg 1b \n"
: "+r"(src_ptr), // %0
"+r"(src_stepx_x4), // %1
"+r"(dst_ptr), // %2
"+rm"(dst_width), // %3
"+r"(src_stepx_x12), // %4
"+r"(row1) // %5
:
: "memory", "cc"
#if defined(__SSE2__)
, "xmm0", "xmm1", "xmm2", "xmm3"
#endif
);
}
// Bilinear row filtering combines 4x2 -> 4x1. SSE2 version
#define HAS_SCALEARGBFILTERROWS_SSE2
static void ScaleARGBFilterRows_SSE2(uint8* dst_ptr,