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

scale to half size use pmadd/pavgw to horizontal averaging.

Browse Source 
BUG=314
TESTED=out\release\libyuv_unittest.exe --gtest_catch_exceptions=0 --gtest_filter=*.ScaleDownBy2*
R=tpsiaki@google.com

Review URL: https://webrtc-codereview.appspot.com/50529004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@1348 16f28f9a-4ce2-e073-06de-1de4eb20be90
This commit is contained in:
fbarchard@google.com 2015-03-27 18:20:21 +00:00
parent e6ca9cc2a2
commit 9ef8999ff3
3 changed files with 15 additions and 11 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: 1347 Version: 1348
License: BSD License: BSD
License File: LICENSE License File: LICENSE

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 1347 #define LIBYUV_VERSION 1348
#endif // INCLUDE_LIBYUV_VERSION_H_ NOLINT #endif // INCLUDE_LIBYUV_VERSION_H_ NOLINT

22
source/scale_win.cc
View File

@ -9,6 +9,7 @@
*/ */
#include "libyuv/row.h" #include "libyuv/row.h"
#include "libyuv/scale_row.h"
#ifdef __cplusplus #ifdef __cplusplus
namespace libyuv { namespace libyuv {
@ -197,6 +198,7 @@ void ScaleRowDown2Box_SSE2(const uint8* src_ptr, ptrdiff_t src_stride,
} }
} }
#ifdef HAS_SCALEROWDOWN2_AVX2
// Blends 64x2 rectangle to 32x1. // Blends 64x2 rectangle to 32x1.
__declspec(naked) __declspec(align(16)) __declspec(naked) __declspec(align(16))
void ScaleRowDown2Box_AVX2(const uint8* src_ptr, ptrdiff_t src_stride, void ScaleRowDown2Box_AVX2(const uint8* src_ptr, ptrdiff_t src_stride,
@ -207,8 +209,11 @@ void ScaleRowDown2Box_AVX2(const uint8* src_ptr, ptrdiff_t src_stride,
mov esi, [esp + 4 + 8] // src_stride mov esi, [esp + 4 + 8] // src_stride
mov edx, [esp + 4 + 12] // dst_ptr mov edx, [esp + 4 + 12] // dst_ptr
mov ecx, [esp + 4 + 16] // dst_width mov ecx, [esp + 4 + 16] // dst_width
vpcmpeqb ymm5, ymm5, ymm5 // generate mask 0x00ff00ff
vpsrlw ymm5, ymm5, 8 vpcmpeqb ymm4, ymm4, ymm4
vpsrlw ymm4, ymm4, 15 // '1' constant, 16b
vpackuswb ymm4, ymm4, ymm4 // '1' constant, 8b
vpxor ymm5, ymm5, ymm5 // constant 0
wloop: wloop:
vmovdqu ymm0, [eax] vmovdqu ymm0, [eax]
@ -217,14 +222,12 @@ void ScaleRowDown2Box_AVX2(const uint8* src_ptr, ptrdiff_t src_stride,
vpavgb ymm1, ymm1, [eax + esi + 32] vpavgb ymm1, ymm1, [eax + esi + 32]
lea eax, [eax + 64] lea eax, [eax + 64]
vpsrlw ymm2, ymm0, 8 // average columns (32 to 16 pixels) vpmaddubsw ymm0, ymm0, ymm4 // add horizontally
vpsrlw ymm3, ymm1, 8 vpmaddubsw ymm1, ymm1, ymm4
vpand ymm0, ymm0, ymm5 vpavgw ymm0, ymm0, ymm5 // (x+1) >> 1
vpand ymm1, ymm1, ymm5 vpavgw ymm1, ymm1, ymm5
vpavgw ymm0, ymm0, ymm2
vpavgw ymm1, ymm1, ymm3
vpackuswb ymm0, ymm0, ymm1 vpackuswb ymm0, ymm0, ymm1
vpermq ymm0, ymm0, 0xd8 // unmutate vpermq ymm0, ymm0, 0xd8 // unmutate vpackuswb
vmovdqu [edx], ymm0 vmovdqu [edx], ymm0
lea edx, [edx + 32] lea edx, [edx + 32]
@ -236,6 +239,7 @@ void ScaleRowDown2Box_AVX2(const uint8* src_ptr, ptrdiff_t src_stride,
ret ret
} }
} }
#endif // HAS_SCALEROWDOWN2_AVX2
// Point samples 32 pixels to 8 pixels. // Point samples 32 pixels to 8 pixels.
__declspec(naked) __declspec(align(16)) __declspec(naked) __declspec(align(16))