coffee/libyuv
1
0
Fork 0
mirror of https://chromium.googlesource.com/libyuv/libyuv synced 2025-12-06 16:56:55 +08:00
Code Issues Packages Projects Releases Wiki Activity

clang use scalewin

Browse Source 
R=harryjin@google.com
TBR=harryjin@google.com
BUG=libyuv:469

Review URL: https://webrtc-codereview.appspot.com/51329004.
This commit is contained in:
Frank Barchard 2015-08-18 14:50:27 -07:00
parent cda9d38a4e
commit 94d4269936
6 changed files with 45 additions and 43 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: 1466 Version: 1467
License: BSD License: BSD
License File: LICENSE License File: LICENSE

6
include/libyuv/rotate_row.h
View File

@ -23,12 +23,6 @@ extern "C" {
#define LIBYUV_DISABLE_X86 #define LIBYUV_DISABLE_X86
#endif #endif
// Visual C 2012 required for AVX2.
#if defined(_M_IX86) && !defined(__clang__) && \
defined(_MSC_VER) && _MSC_VER >= 1700
#define VISUALC_HAS_AVX2 1
#endif // VisualStudio >= 2012
// The following are available for Visual C and clangcl 32 bit: // The following are available for Visual C and clangcl 32 bit:
#if !defined(LIBYUV_DISABLE_X86) && defined(_M_IX86) #if !defined(LIBYUV_DISABLE_X86) && defined(_M_IX86)
#define HAS_TRANSPOSEWX8_SSSE3 #define HAS_TRANSPOSEWX8_SSSE3

14
include/libyuv/scale_row.h
View File

@ -30,6 +30,13 @@ extern "C" {
#define VISUALC_HAS_AVX2 1 #define VISUALC_HAS_AVX2 1
#endif // VisualStudio >= 2012 #endif // VisualStudio >= 2012
// clang >= 3.4.0 required for AVX2.
#if defined(__clang__) && (defined(__x86_64__) || defined(__i386__))
#if (__clang_major__ > 3) || (__clang_major__ == 3 && (__clang_minor__ >= 4))
#define CLANG_HAS_AVX2 1
#endif // clang >= 3.4
#endif // __clang__
// The following are available on all x86 platforms: // The following are available on all x86 platforms:
#if !defined(LIBYUV_DISABLE_X86) && \ #if !defined(LIBYUV_DISABLE_X86) && \
(defined(_M_IX86) || defined(__x86_64__) || defined(__i386__)) (defined(_M_IX86) || defined(__x86_64__) || defined(__i386__))
@ -48,15 +55,16 @@ extern "C" {
#define HAS_SCALEROWDOWN4_SSE2 #define HAS_SCALEROWDOWN4_SSE2
#endif #endif
// The following are available on VS2012: // The following are available for Visual C and clangcl 32 bit:
#if !defined(LIBYUV_DISABLE_X86) && defined(VISUALC_HAS_AVX2) #if !defined(LIBYUV_DISABLE_X86) && defined(_M_IX86) && \
(defined(VISUALC_HAS_AVX2) || defined(CLANG_HAS_AVX2))
#define HAS_SCALEADDROW_AVX2 #define HAS_SCALEADDROW_AVX2
#define HAS_SCALEROWDOWN2_AVX2 #define HAS_SCALEROWDOWN2_AVX2
#define HAS_SCALEROWDOWN4_AVX2 #define HAS_SCALEROWDOWN4_AVX2
#endif #endif
// The following are available on Visual C: // The following are available on Visual C:
#if !defined(LIBYUV_DISABLE_X86) && defined(_M_IX86) && !defined(__clang__) #if !defined(LIBYUV_DISABLE_X86) && defined(_M_IX86)
#define HAS_SCALEADDROW_SSE2 #define HAS_SCALEADDROW_SSE2
#endif #endif

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 1466 #define LIBYUV_VERSION 1467
#endif // INCLUDE_LIBYUV_VERSION_H_ NOLINT #endif // INCLUDE_LIBYUV_VERSION_H_ NOLINT

3
source/scale_gcc.cc
View File

@ -16,7 +16,8 @@ extern "C" {
#endif #endif
// This module is for GCC x86 and x64. // This module is for GCC x86 and x64.
#if !defined(LIBYUV_DISABLE_X86) && (defined(__x86_64__) || defined(__i386__)) #if !defined(LIBYUV_DISABLE_X86) && \
(defined(__x86_64__) || (defined(__i386__) && !defined(_MSC_VER)))
// Offsets for source bytes 0 to 9 // Offsets for source bytes 0 to 9
static uvec8 kShuf0 = static uvec8 kShuf0 =

61
source/scale_win.cc
View File

@ -16,9 +16,8 @@ namespace libyuv {
extern "C" { extern "C" {
#endif #endif
// This module is for Visual C x86. // This module is for 32 bit Visual C x86 and clangcl
#if !defined(LIBYUV_DISABLE_X86) && defined(_M_IX86) && \ #if !defined(LIBYUV_DISABLE_X86) && defined(_M_IX86)
defined(_MSC_VER) && !defined(__clang__)
// Offsets for source bytes 0 to 9 // Offsets for source bytes 0 to 9
static uvec8 kShuf0 = static uvec8 kShuf0 =
@ -499,9 +498,9 @@ void ScaleRowDown34_SSSE3(const uint8* src_ptr, ptrdiff_t src_stride,
// src_stride ignored // src_stride ignored
mov edx, [esp + 12] // dst_ptr mov edx, [esp + 12] // dst_ptr
mov ecx, [esp + 16] // dst_width mov ecx, [esp + 16] // dst_width
movdqa xmm3, kShuf0 movdqa xmm3, xmmword ptr kShuf0
movdqa xmm4, kShuf1 movdqa xmm4, xmmword ptr kShuf1
movdqa xmm5, kShuf2 movdqa xmm5, xmmword ptr kShuf2
wloop: wloop:
movdqu xmm0, [eax] movdqu xmm0, [eax]
@ -548,12 +547,12 @@ void ScaleRowDown34_1_Box_SSSE3(const uint8* src_ptr,
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
movdqa xmm2, kShuf01 movdqa xmm2, xmmword ptr kShuf01
movdqa xmm3, kShuf11 movdqa xmm3, xmmword ptr kShuf11
movdqa xmm4, kShuf21 movdqa xmm4, xmmword ptr kShuf21
movdqa xmm5, kMadd01 movdqa xmm5, xmmword ptr kMadd01
movdqa xmm6, kMadd11 movdqa xmm6, xmmword ptr kMadd11
movdqa xmm7, kRound34 movdqa xmm7, xmmword ptr kRound34
wloop: wloop:
movdqu xmm0, [eax] // pixels 0..7 movdqu xmm0, [eax] // pixels 0..7
@ -579,7 +578,7 @@ void ScaleRowDown34_1_Box_SSSE3(const uint8* src_ptr,
lea eax, [eax + 32] lea eax, [eax + 32]
pavgb xmm0, xmm1 pavgb xmm0, xmm1
pshufb xmm0, xmm4 pshufb xmm0, xmm4
movdqa xmm1, kMadd21 movdqa xmm1, xmmword ptr kMadd21
pmaddubsw xmm0, xmm1 pmaddubsw xmm0, xmm1
paddsw xmm0, xmm7 paddsw xmm0, xmm7
psrlw xmm0, 2 psrlw xmm0, 2
@ -605,12 +604,12 @@ void ScaleRowDown34_0_Box_SSSE3(const uint8* src_ptr,
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
movdqa xmm2, kShuf01 movdqa xmm2, xmmword ptr kShuf01
movdqa xmm3, kShuf11 movdqa xmm3, xmmword ptr kShuf11
movdqa xmm4, kShuf21 movdqa xmm4, xmmword ptr kShuf21
movdqa xmm5, kMadd01 movdqa xmm5, xmmword ptr kMadd01
movdqa xmm6, kMadd11 movdqa xmm6, xmmword ptr kMadd11
movdqa xmm7, kRound34 movdqa xmm7, xmmword ptr kRound34
wloop: wloop:
movdqu xmm0, [eax] // pixels 0..7 movdqu xmm0, [eax] // pixels 0..7
@ -639,7 +638,7 @@ void ScaleRowDown34_0_Box_SSSE3(const uint8* src_ptr,
pavgb xmm1, xmm0 pavgb xmm1, xmm0
pavgb xmm0, xmm1 pavgb xmm0, xmm1
pshufb xmm0, xmm4 pshufb xmm0, xmm4
movdqa xmm1, kMadd21 movdqa xmm1, xmmword ptr kMadd21
pmaddubsw xmm0, xmm1 pmaddubsw xmm0, xmm1
paddsw xmm0, xmm7 paddsw xmm0, xmm7
psrlw xmm0, 2 psrlw xmm0, 2
@ -665,8 +664,8 @@ void ScaleRowDown38_SSSE3(const uint8* src_ptr, ptrdiff_t src_stride,
// src_stride ignored // src_stride ignored
mov edx, [esp + 12] // dst_ptr mov edx, [esp + 12] // dst_ptr
mov ecx, [esp + 16] // dst_width mov ecx, [esp + 16] // dst_width
movdqa xmm4, kShuf38a movdqa xmm4, xmmword ptr kShuf38a
movdqa xmm5, kShuf38b movdqa xmm5, xmmword ptr kShuf38b
xloop: xloop:
movdqu xmm0, [eax] // 16 pixels -> 0,1,2,3,4,5 movdqu xmm0, [eax] // 16 pixels -> 0,1,2,3,4,5
@ -698,9 +697,9 @@ void ScaleRowDown38_3_Box_SSSE3(const uint8* src_ptr,
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
movdqa xmm2, kShufAc movdqa xmm2, xmmword ptr kShufAc
movdqa xmm3, kShufAc3 movdqa xmm3, xmmword ptr kShufAc3
movdqa xmm4, kScaleAc33 movdqa xmm4, xmmword ptr kScaleAc33
pxor xmm5, xmm5 pxor xmm5, xmm5
xloop: xloop:
@ -763,10 +762,10 @@ void ScaleRowDown38_2_Box_SSSE3(const uint8* src_ptr,
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
movdqa xmm2, kShufAb0 movdqa xmm2, xmmword ptr kShufAb0
movdqa xmm3, kShufAb1 movdqa xmm3, xmmword ptr kShufAb1
movdqa xmm4, kShufAb2 movdqa xmm4, xmmword ptr kShufAb2
movdqa xmm5, kScaleAb2 movdqa xmm5, xmmword ptr kScaleAb2
xloop: xloop:
movdqu xmm0, [eax] // average 2 rows into xmm0 movdqu xmm0, [eax] // average 2 rows into xmm0
@ -1233,8 +1232,8 @@ void ScaleARGBFilterCols_SSSE3(uint8* dst_argb, const uint8* src_argb,
mov ecx, [esp + 8 + 12] // dst_width mov ecx, [esp + 8 + 12] // dst_width
movd xmm2, [esp + 8 + 16] // x movd xmm2, [esp + 8 + 16] // x
movd xmm3, [esp + 8 + 20] // dx movd xmm3, [esp + 8 + 20] // dx
movdqa xmm4, kShuffleColARGB movdqa xmm4, xmmword ptr kShuffleColARGB
movdqa xmm5, kShuffleFractions movdqa xmm5, xmmword ptr kShuffleFractions
pcmpeqb xmm6, xmm6 // generate 0x007f for inverting fraction. pcmpeqb xmm6, xmm6 // generate 0x007f for inverting fraction.
psrlw xmm6, 9 psrlw xmm6, 9
pextrw eax, xmm2, 1 // get x0 integer. preroll pextrw eax, xmm2, 1 // get x0 integer. preroll