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

lint fix test

Browse Source 
BUG=none
TEST=gcl lint lintfix2
Review URL: https://webrtc-codereview.appspot.com/458003

git-svn-id: http://libyuv.googlecode.com/svn/trunk@220 16f28f9a-4ce2-e073-06de-1de4eb20be90
This commit is contained in:
fbarchard@google.com 2012-03-22 22:36:44 +00:00
parent 8670b1ae04
commit 2b9c210803
17 changed files with 161 additions and 174 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: 219 Version: 220
License: BSD License: BSD
License File: LICENSE License File: LICENSE

2
include/libyuv/version.h
View File

@ -11,7 +11,7 @@
#ifndef INCLUDE_LIBYUV_VERSION_H_ #ifndef INCLUDE_LIBYUV_VERSION_H_
#define INCLUDE_LIBYUV_VERSION_H_ #define INCLUDE_LIBYUV_VERSION_H_
#define LIBYUV_VERSION 219 #define LIBYUV_VERSION 220
#endif // INCLUDE_LIBYUV_VERSION_H_ #endif // INCLUDE_LIBYUV_VERSION_H_

7
include/libyuv/video_common.h
View File

@ -25,6 +25,7 @@ extern "C" {
////////////////////////////////////////////////////////////////////////////// //////////////////////////////////////////////////////////////////////////////
// Definition of FourCC codes // Definition of FourCC codes
////////////////////////////////////////////////////////////////////////////// //////////////////////////////////////////////////////////////////////////////
// Convert four characters to a FourCC code. // Convert four characters to a FourCC code.
// Needs to be a macro otherwise the OS X compiler complains when the kFormat* // Needs to be a macro otherwise the OS X compiler complains when the kFormat*
// constants are used in a switch. // constants are used in a switch.
@ -57,9 +58,9 @@ enum FourCC {
FOURCC_ABGR = FOURCC('A', 'B', 'G', 'R'), FOURCC_ABGR = FOURCC('A', 'B', 'G', 'R'),
FOURCC_BGRA = FOURCC('B', 'G', 'R', 'A'), FOURCC_BGRA = FOURCC('B', 'G', 'R', 'A'),
FOURCC_ARGB = FOURCC('A', 'R', 'G', 'B'), FOURCC_ARGB = FOURCC('A', 'R', 'G', 'B'),
FOURCC_RGBP = FOURCC('R', 'G', 'B', 'P'), // bgr565 FOURCC_RGBP = FOURCC('R', 'G', 'B', 'P'), // bgr565
FOURCC_RGBO = FOURCC('R', 'G', 'B', 'O'), // abgr1555 FOURCC_RGBO = FOURCC('R', 'G', 'B', 'O'), // abgr1555
FOURCC_R444 = FOURCC('R', '4', '4', '4'), // argb4444 FOURCC_R444 = FOURCC('R', '4', '4', '4'), // argb4444
FOURCC_MJPG = FOURCC('M', 'J', 'P', 'G'), FOURCC_MJPG = FOURCC('M', 'J', 'P', 'G'),
FOURCC_RAW = FOURCC('r', 'a', 'w', ' '), FOURCC_RAW = FOURCC('r', 'a', 'w', ' '),
FOURCC_NV21 = FOURCC('N', 'V', '2', '1'), FOURCC_NV21 = FOURCC('N', 'V', '2', '1'),

9
source/compare.cc
View File

@ -18,7 +18,7 @@
#include "libyuv/basic_types.h" #include "libyuv/basic_types.h"
#include "libyuv/cpu_id.h" #include "libyuv/cpu_id.h"
#include "row.h" #include "source/row.h"
#ifdef __cplusplus #ifdef __cplusplus
namespace libyuv { namespace libyuv {
@ -42,7 +42,7 @@ uint32 HashDjb2(const uint8* src, uint64 count, uint32 seed) {
static uint32 SumSquareError_NEON(const uint8* src_a, const uint8* src_b, static uint32 SumSquareError_NEON(const uint8* src_a, const uint8* src_b,
int count) { int count) {
volatile uint32 sse; volatile uint32 sse;
asm volatile ( asm volatile(
"vmov.u8 q7, #0 \n" "vmov.u8 q7, #0 \n"
"vmov.u8 q9, #0 \n" "vmov.u8 q9, #0 \n"
"vmov.u8 q8, #0 \n" "vmov.u8 q8, #0 \n"
@ -71,8 +71,7 @@ static uint32 SumSquareError_NEON(const uint8* src_a, const uint8* src_b,
"+r"(count), "+r"(count),
"=r"(sse) "=r"(sse)
: :
: "memory", "cc", "q0", "q1", "q2", "q3", "q7", "q8", "q9", "q10" : "memory", "cc", "q0", "q1", "q2", "q3", "q7", "q8", "q9", "q10");
);
return sse; return sse;
} }
@ -122,7 +121,7 @@ static uint32 SumSquareError_SSE2(const uint8* src_a, const uint8* src_b,
static uint32 SumSquareError_SSE2(const uint8* src_a, const uint8* src_b, static uint32 SumSquareError_SSE2(const uint8* src_a, const uint8* src_b,
int count) { int count) {
uint32 sse; uint32 sse;
asm volatile ( asm volatile(
"pxor %%xmm0,%%xmm0 \n" "pxor %%xmm0,%%xmm0 \n"
"pxor %%xmm5,%%xmm5 \n" "pxor %%xmm5,%%xmm5 \n"
"sub %0,%1 \n" "sub %0,%1 \n"

72
source/convert.cc
View File

@ -19,7 +19,7 @@
#include "libyuv/planar_functions.h" #include "libyuv/planar_functions.h"
#include "libyuv/rotate.h" #include "libyuv/rotate.h"
#include "libyuv/video_common.h" #include "libyuv/video_common.h"
#include "row.h" #include "source/row.h"
#ifdef __cplusplus #ifdef __cplusplus
namespace libyuv { namespace libyuv {
@ -78,7 +78,7 @@ static void HalfRow_SSE2(const uint8* src_uv, int src_uv_stride,
movdqa xmm0, [eax] movdqa xmm0, [eax]
pavgb xmm0, [eax + edx] pavgb xmm0, [eax + edx]
sub ecx, 16 sub ecx, 16
movdqa [eax + edi], xmm0 movdqa [eax + edi], xmm0 // NOLINT
lea eax, [eax + 16] lea eax, [eax + 16]
jg convertloop jg convertloop
pop edi pop edi
@ -86,11 +86,11 @@ static void HalfRow_SSE2(const uint8* src_uv, int src_uv_stride,
} }
} }
#elif (defined(__x86_64__) || defined(__i386__)) && !defined(YUV_DISABLE_ASM) #elif defined(__x86_64__) || defined(__i386__) && !defined(YUV_DISABLE_ASM)
#define HAS_HALFROW_SSE2 #define HAS_HALFROW_SSE2
static void HalfRow_SSE2(const uint8* src_uv, int src_uv_stride, static void HalfRow_SSE2(const uint8* src_uv, int src_uv_stride,
uint8* dst_uv, int pix) { uint8* dst_uv, int pix) {
asm volatile ( asm volatile(
"sub %0,%1 \n" "sub %0,%1 \n"
"1: \n" "1: \n"
"movdqa (%0),%%xmm0 \n" "movdqa (%0),%%xmm0 \n"
@ -137,7 +137,7 @@ int I422ToI420(const uint8* src_y, int src_stride_y,
} }
int halfwidth = (width + 1) >> 1; int halfwidth = (width + 1) >> 1;
void (*HalfRow)(const uint8* src_uv, int src_uv_stride, void (*HalfRow)(const uint8* src_uv, int src_uv_stride,
uint8* dst_uv, int pix); uint8* dst_uv, int pix) = HalfRow_C;
#if defined(HAS_HALFROW_SSE2) #if defined(HAS_HALFROW_SSE2)
if (TestCpuFlag(kCpuHasSSE2) && if (TestCpuFlag(kCpuHasSSE2) &&
IS_ALIGNED(halfwidth, 16) && IS_ALIGNED(halfwidth, 16) &&
@ -146,11 +146,8 @@ int I422ToI420(const uint8* src_y, int src_stride_y,
IS_ALIGNED(dst_u, 16) && IS_ALIGNED(dst_stride_u, 16) && IS_ALIGNED(dst_u, 16) && IS_ALIGNED(dst_stride_u, 16) &&
IS_ALIGNED(dst_v, 16) && IS_ALIGNED(dst_stride_v, 16)) { IS_ALIGNED(dst_v, 16) && IS_ALIGNED(dst_stride_v, 16)) {
HalfRow = HalfRow_SSE2; HalfRow = HalfRow_SSE2;
} else
#endif
{
HalfRow = HalfRow_C;
} }
#endif
// Copy Y plane // Copy Y plane
if (dst_y) { if (dst_y) {
@ -186,7 +183,7 @@ int I422ToI420(const uint8* src_y, int src_stride_y,
#define HAS_SCALEROWDOWN2_NEON #define HAS_SCALEROWDOWN2_NEON
void ScaleRowDown2Int_NEON(const uint8* src_ptr, int src_stride, void ScaleRowDown2Int_NEON(const uint8* src_ptr, int src_stride,
uint8* dst, int dst_width); uint8* dst, int dst_width);
#elif (defined(_M_IX86) || defined(__x86_64__) || defined(__i386__)) && \ #elif defined(_M_IX86) || defined(__x86_64__) || defined(__i386__) && \
!defined(YUV_DISABLE_ASM) !defined(YUV_DISABLE_ASM)
void ScaleRowDown2Int_SSE2(const uint8* src_ptr, int src_stride, void ScaleRowDown2Int_SSE2(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width); uint8* dst_ptr, int dst_width);
@ -213,14 +210,13 @@ int I444ToI420(const uint8* src_y, int src_stride_y,
} }
int halfwidth = (width + 1) >> 1; int halfwidth = (width + 1) >> 1;
void (*ScaleRowDown2)(const uint8* src_ptr, int src_stride, void (*ScaleRowDown2)(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width); uint8* dst_ptr, int dst_width) = ScaleRowDown2Int_C;
#if defined(HAS_SCALEROWDOWN2_NEON) #if defined(HAS_SCALEROWDOWN2_NEON)
if (TestCpuFlag(kCpuHasNEON) && if (TestCpuFlag(kCpuHasNEON) &&
IS_ALIGNED(halfwidth, 16)) { IS_ALIGNED(halfwidth, 16)) {
ScaleRowDown2 = ScaleRowDown2Int_NEON; ScaleRowDown2 = ScaleRowDown2Int_NEON;
} else }
#endif #elif defined(HAS_SCALEROWDOWN2_SSE2)
#if defined(HAS_SCALEROWDOWN2_SSE2)
if (TestCpuFlag(kCpuHasSSE2) && if (TestCpuFlag(kCpuHasSSE2) &&
IS_ALIGNED(halfwidth, 16) && IS_ALIGNED(halfwidth, 16) &&
IS_ALIGNED(src_u, 16) && IS_ALIGNED(src_stride_u, 16) && IS_ALIGNED(src_u, 16) && IS_ALIGNED(src_stride_u, 16) &&
@ -228,10 +224,8 @@ int I444ToI420(const uint8* src_y, int src_stride_y,
IS_ALIGNED(dst_u, 16) && IS_ALIGNED(dst_stride_u, 16) && IS_ALIGNED(dst_u, 16) && IS_ALIGNED(dst_stride_u, 16) &&
IS_ALIGNED(dst_v, 16) && IS_ALIGNED(dst_stride_v, 16)) { IS_ALIGNED(dst_v, 16) && IS_ALIGNED(dst_stride_v, 16)) {
ScaleRowDown2 = ScaleRowDown2Int_SSE2; ScaleRowDown2 = ScaleRowDown2Int_SSE2;
#endif
{
ScaleRowDown2 = ScaleRowDown2Int_C;
} }
#endif
// Copy Y plane // Copy Y plane
if (dst_y) { if (dst_y) {
@ -395,11 +389,12 @@ static int X420ToI420(const uint8* src_y,
} }
int halfwidth = (width + 1) >> 1; int halfwidth = (width + 1) >> 1;
void (*SplitUV)(const uint8* src_uv, uint8* dst_u, uint8* dst_v, int pix); void (*SplitUV)(const uint8* src_uv, uint8* dst_u, uint8* dst_v, int pix) =
SplitUV_C;
#if defined(HAS_SPLITUV_NEON) #if defined(HAS_SPLITUV_NEON)
if (TestCpuFlag(kCpuHasNEON) && IS_ALIGNED(halfwidth, 16)) { if (TestCpuFlag(kCpuHasNEON) && IS_ALIGNED(halfwidth, 16)) {
SplitUV = SplitUV_NEON; SplitUV = SplitUV_NEON;
} else }
#elif defined(HAS_SPLITUV_SSE2) #elif defined(HAS_SPLITUV_SSE2)
if (TestCpuFlag(kCpuHasSSE2) && if (TestCpuFlag(kCpuHasSSE2) &&
IS_ALIGNED(halfwidth, 16) && IS_ALIGNED(halfwidth, 16) &&
@ -407,11 +402,8 @@ static int X420ToI420(const uint8* src_y,
IS_ALIGNED(dst_u, 16) && IS_ALIGNED(dst_stride_u, 16) && IS_ALIGNED(dst_u, 16) && IS_ALIGNED(dst_stride_u, 16) &&
IS_ALIGNED(dst_v, 16) && IS_ALIGNED(dst_stride_v, 16)) { IS_ALIGNED(dst_v, 16) && IS_ALIGNED(dst_stride_v, 16)) {
SplitUV = SplitUV_SSE2; SplitUV = SplitUV_SSE2;
} else
#endif
{
SplitUV = SplitUV_C;
} }
#endif
if (dst_y) { if (dst_y) {
CopyPlane2(src_y, src_stride_y0, src_stride_y1, dst_y, dst_stride_y, CopyPlane2(src_y, src_stride_y0, src_stride_y1, dst_y, dst_stride_y,
@ -483,7 +475,7 @@ static void SplitYUY2_SSE2(const uint8* src_yuy2,
pand xmm2, xmm5 // even bytes are Y pand xmm2, xmm5 // even bytes are Y
pand xmm3, xmm5 pand xmm3, xmm5
packuswb xmm2, xmm3 packuswb xmm2, xmm3
movdqa [edx], xmm2 movdqa [edx], xmm2 // NOLINT
lea edx, [edx + 16] lea edx, [edx + 16]
psrlw xmm0, 8 // YUYV -> UVUV psrlw xmm0, 8 // YUYV -> UVUV
psrlw xmm1, 8 psrlw xmm1, 8
@ -491,12 +483,12 @@ static void SplitYUY2_SSE2(const uint8* src_yuy2,
movdqa xmm1, xmm0 movdqa xmm1, xmm0
pand xmm0, xmm5 // U pand xmm0, xmm5 // U
packuswb xmm0, xmm0 packuswb xmm0, xmm0
movq qword ptr [esi], xmm0 movq qword ptr [esi], xmm0 // NOLINT
lea esi, [esi + 8] lea esi, [esi + 8]
psrlw xmm1, 8 // V psrlw xmm1, 8 // V
packuswb xmm1, xmm1 packuswb xmm1, xmm1
sub ecx, 16 sub ecx, 16
movq qword ptr [edi], xmm1 movq qword ptr [edi], xmm1 // NOLINT
lea edi, [edi + 8] lea edi, [edi + 8]
jg convertloop jg convertloop
@ -506,11 +498,11 @@ static void SplitYUY2_SSE2(const uint8* src_yuy2,
} }
} }
#elif (defined(__x86_64__) || defined(__i386__)) && !defined(YUV_DISABLE_ASM) #elif defined(__x86_64__) || defined(__i386__) && !defined(YUV_DISABLE_ASM)
#define HAS_SPLITYUY2_SSE2 #define HAS_SPLITYUY2_SSE2
static void SplitYUY2_SSE2(const uint8* src_yuy2, uint8* dst_y, static void SplitYUY2_SSE2(const uint8* src_yuy2, uint8* dst_y,
uint8* dst_u, uint8* dst_v, int pix) { uint8* dst_u, uint8* dst_v, int pix) {
asm volatile ( asm volatile(
"pcmpeqb %%xmm5,%%xmm5 \n" "pcmpeqb %%xmm5,%%xmm5 \n"
"psrlw $0x8,%%xmm5 \n" "psrlw $0x8,%%xmm5 \n"
"1: \n" "1: \n"
@ -604,19 +596,17 @@ int Q420ToI420(const uint8* src_y, int src_stride_y,
} }
#endif #endif
void (*SplitYUY2)(const uint8* src_yuy2, void (*SplitYUY2)(const uint8* src_yuy2, uint8* dst_y, uint8* dst_u,
uint8* dst_y, uint8* dst_u, uint8* dst_v, int pix); uint8* dst_v, int pix) = SplitYUY2_C;
#if defined(HAS_SPLITYUY2_SSE2) #if defined(HAS_SPLITYUY2_SSE2)
if (TestCpuFlag(kCpuHasSSE2) && if (TestCpuFlag(kCpuHasSSE2) &&
IS_ALIGNED(width, 16) && IS_ALIGNED(width, 16) &&
IS_ALIGNED(src_yuy2, 16) && IS_ALIGNED(src_stride_yuy2, 16) && IS_ALIGNED(src_yuy2, 16) && IS_ALIGNED(src_stride_yuy2, 16) &&
IS_ALIGNED(dst_y, 16) && IS_ALIGNED(dst_stride_y, 16)) { IS_ALIGNED(dst_y, 16) && IS_ALIGNED(dst_stride_y, 16)) {
SplitYUY2 = SplitYUY2_SSE2; SplitYUY2 = SplitYUY2_SSE2;
} else
#endif
{
SplitYUY2 = SplitYUY2_C;
} }
#endif
for (int y = 0; y < height; y += 2) { for (int y = 0; y < height; y += 2) {
CopyRow(src_y, dst_y, width); CopyRow(src_y, dst_y, width);
dst_y += dst_stride_y; dst_y += dst_stride_y;
@ -800,13 +790,13 @@ int UYVYToI420(const uint8* src_uyvy, int src_stride_uyvy,
#endif #endif
#ifdef LIBYUV_LITTLE_ENDIAN #ifdef LIBYUV_LITTLE_ENDIAN
#define READWORD(p) (*((uint32*) (p))) #define READWORD(p) (*reinterpret_cast<const uint32*>(p))
#else #else
static inline uint32 READWORD(const uint8* p) { static inline uint32 READWORD(const uint8* p) {
return (uint32) p[0] | return static_cast<uint32>(p[0]) |
((uint32) (p[1]) << 8) | (static_cast<uint32>(p[1]) << 8) |
((uint32) (p[2]) << 16) | (static_cast<uint32>(p[2]) << 16) |
((uint32) (p[3]) << 24); (static_cast<uint32>(p[3]) << 24);
} }
#endif #endif
@ -1599,7 +1589,7 @@ int ConvertToI420(const uint8* sample, size_t sample_size,
int tmp_y_stride = y_stride; int tmp_y_stride = y_stride;
int tmp_u_stride = u_stride; int tmp_u_stride = u_stride;
int tmp_v_stride = v_stride; int tmp_v_stride = v_stride;
uint8* buf = 0; uint8* buf = NULL;
int abs_dst_height = (dst_height < 0) ? -dst_height : dst_height; int abs_dst_height = (dst_height < 0) ? -dst_height : dst_height;
if (need_rot) { if (need_rot) {
int y_size = dst_width * abs_dst_height; int y_size = dst_width * abs_dst_height;
@ -1618,7 +1608,7 @@ int ConvertToI420(const uint8* sample, size_t sample_size,
switch (format) { switch (format) {
// Single plane formats // Single plane formats
case FOURCC_YUY2: case FOURCC_YUY2:
src = sample + (aligned_src_width * crop_y + crop_x) * 2 ; src = sample + (aligned_src_width * crop_y + crop_x) * 2;
r = YUY2ToI420(src, aligned_src_width * 2, r = YUY2ToI420(src, aligned_src_width * 2,
y, y_stride, y, y_stride,
u, u_stride, u, u_stride,

6
source/convert_from.cc
View File

@ -17,7 +17,7 @@
#include "libyuv/planar_functions.h" #include "libyuv/planar_functions.h"
#include "libyuv/rotate.h" #include "libyuv/rotate.h"
#include "libyuv/video_common.h" #include "libyuv/video_common.h"
#include "row.h" #include "source/row.h"
#ifdef __cplusplus #ifdef __cplusplus
namespace libyuv { namespace libyuv {
@ -289,7 +289,7 @@ static void I42xToYUY2Row_SSE2(const uint8* src_y,
const uint8* src_u, const uint8* src_u,
const uint8* src_v, const uint8* src_v,
uint8* dst_frame, int width) { uint8* dst_frame, int width) {
asm volatile ( asm volatile(
"sub %1,%2 \n" "sub %1,%2 \n"
"1: \n" "1: \n"
"movq (%1),%%xmm2 \n" "movq (%1),%%xmm2 \n"
@ -324,7 +324,7 @@ static void I42xToUYVYRow_SSE2(const uint8* src_y,
const uint8* src_u, const uint8* src_u,
const uint8* src_v, const uint8* src_v,
uint8* dst_frame, int width) { uint8* dst_frame, int width) {
asm volatile ( asm volatile(
"sub %1,%2 \n" "sub %1,%2 \n"
"1: \n" "1: \n"
"movq (%1),%%xmm2 \n" "movq (%1),%%xmm2 \n"

10
source/cpu_id.cc
View File

@ -24,21 +24,19 @@
// TODO(fbarchard): Use cpuid.h when gcc 4.4 is used on OSX and Linux. // TODO(fbarchard): Use cpuid.h when gcc 4.4 is used on OSX and Linux.
#if (defined(__pic__) || defined(__APPLE__)) && defined(__i386__) #if (defined(__pic__) || defined(__APPLE__)) && defined(__i386__)
static __inline void __cpuid(int cpu_info[4], int info_type) { static __inline void __cpuid(int cpu_info[4], int info_type) {
asm volatile ( asm volatile(
"mov %%ebx, %%edi \n" "mov %%ebx, %%edi \n"
"cpuid \n" "cpuid \n"
"xchg %%edi, %%ebx \n" "xchg %%edi, %%ebx \n"
: "=a"(cpu_info[0]), "=D"(cpu_info[1]), "=c"(cpu_info[2]), "=d"(cpu_info[3]) : "=a"(cpu_info[0]), "=D"(cpu_info[1]), "=c"(cpu_info[2]), "=d"(cpu_info[3])
: "a"(info_type) : "a"(info_type));
);
} }
#elif defined(__i386__) || defined(__x86_64__) #elif defined(__i386__) || defined(__x86_64__)
static __inline void __cpuid(int cpu_info[4], int info_type) { static __inline void __cpuid(int cpu_info[4], int info_type) {
asm volatile ( asm volatile(
"cpuid \n" "cpuid \n"
: "=a"(cpu_info[0]), "=b"(cpu_info[1]), "=c"(cpu_info[2]), "=d"(cpu_info[3]) : "=a"(cpu_info[0]), "=b"(cpu_info[1]), "=c"(cpu_info[2]), "=d"(cpu_info[3])
: "a"(info_type) : "a"(info_type));
);
} }
#endif #endif

4
source/format_conversion.cc
View File

@ -13,7 +13,7 @@
#include "libyuv/basic_types.h" #include "libyuv/basic_types.h"
#include "libyuv/cpu_id.h" #include "libyuv/cpu_id.h"
#include "libyuv/video_common.h" #include "libyuv/video_common.h"
#include "row.h" #include "source/row.h"
#ifdef __cplusplus #ifdef __cplusplus
namespace libyuv { namespace libyuv {
@ -53,7 +53,7 @@ static void ARGBToBayerRow_SSSE3(const uint8* src_argb,
#define HAS_ARGBTOBAYERROW_SSSE3 #define HAS_ARGBTOBAYERROW_SSSE3
static void ARGBToBayerRow_SSSE3(const uint8* src_argb, uint8* dst_bayer, static void ARGBToBayerRow_SSSE3(const uint8* src_argb, uint8* dst_bayer,
uint32 selector, int pix) { uint32 selector, int pix) {
asm volatile ( asm volatile(
"movd %3,%%xmm5 \n" "movd %3,%%xmm5 \n"
"pshufd $0x0,%%xmm5,%%xmm5 \n" "pshufd $0x0,%%xmm5,%%xmm5 \n"
"1: \n" "1: \n"

8
source/planar_functions.cc
View File

@ -13,7 +13,7 @@
#include <string.h> // for memset() #include <string.h> // for memset()
#include "libyuv/cpu_id.h" #include "libyuv/cpu_id.h"
#include "row.h" #include "source/row.h"
#ifdef __cplusplus #ifdef __cplusplus
namespace libyuv { namespace libyuv {
@ -693,7 +693,7 @@ int NV12ToRGB565(const uint8* src_y, int src_stride_y,
#if defined(__ARM_NEON__) && !defined(YUV_DISABLE_ASM) #if defined(__ARM_NEON__) && !defined(YUV_DISABLE_ASM)
#define HAS_SETROW_NEON #define HAS_SETROW_NEON
static void SetRow8_NEON(uint8* dst, uint32 v32, int count) { static void SetRow8_NEON(uint8* dst, uint32 v32, int count) {
asm volatile ( asm volatile(
"vdup.u32 q0, %2 \n" // duplicate 4 ints "vdup.u32 q0, %2 \n" // duplicate 4 ints
"1: \n" "1: \n"
"subs %1, %1, #16 \n" // 16 bytes per loop "subs %1, %1, #16 \n" // 16 bytes per loop
@ -763,7 +763,7 @@ static void SetRows32_X86(uint8* dst, uint32 v32, int width,
#define HAS_SETROW_X86 #define HAS_SETROW_X86
static void SetRow8_X86(uint8* dst, uint32 v32, int width) { static void SetRow8_X86(uint8* dst, uint32 v32, int width) {
size_t width_tmp = static_cast<size_t>(width); size_t width_tmp = static_cast<size_t>(width);
asm volatile ( asm volatile(
"shr $0x2,%1 \n" "shr $0x2,%1 \n"
"rep stosl \n" "rep stosl \n"
: "+D"(dst), // %0 : "+D"(dst), // %0
@ -778,7 +778,7 @@ static void SetRows32_X86(uint8* dst, uint32 v32, int width,
for (int y = 0; y < height; ++y) { for (int y = 0; y < height; ++y) {
size_t width_tmp = static_cast<size_t>(width); size_t width_tmp = static_cast<size_t>(width);
uint32* d = reinterpret_cast<uint32*>(dst); uint32* d = reinterpret_cast<uint32*>(dst);
asm volatile ( asm volatile(
"rep stosl \n" "rep stosl \n"
: "+D"(d), // %0 : "+D"(d), // %0
"+c"(width_tmp) // %1 "+c"(width_tmp) // %1

10
source/rotate.cc
View File

@ -13,8 +13,8 @@
#include "libyuv/cpu_id.h" #include "libyuv/cpu_id.h"
#include "libyuv/convert.h" #include "libyuv/convert.h"
#include "libyuv/planar_functions.h" #include "libyuv/planar_functions.h"
#include "rotate_priv.h" #include "source/rotate_priv.h"
#include "row.h" #include "source/row.h"
#ifdef __cplusplus #ifdef __cplusplus
namespace libyuv { namespace libyuv {
@ -295,7 +295,7 @@ static void TransposeUVWx8_SSE2(const uint8* src, int src_stride,
#define HAS_TRANSPOSE_WX8_SSSE3 #define HAS_TRANSPOSE_WX8_SSSE3
static void TransposeWx8_SSSE3(const uint8* src, int src_stride, static void TransposeWx8_SSSE3(const uint8* src, int src_stride,
uint8* dst, int dst_stride, int width) { uint8* dst, int dst_stride, int width) {
asm volatile ( asm volatile(
// Read in the data from the source pointer. // Read in the data from the source pointer.
// First round of bit swap. // First round of bit swap.
"1: \n" "1: \n"
@ -506,7 +506,7 @@ extern "C" void TransposeUVWx8_SSE2(const uint8* src, int src_stride,
#define HAS_TRANSPOSE_WX8_FAST_SSSE3 #define HAS_TRANSPOSE_WX8_FAST_SSSE3
static void TransposeWx8_FAST_SSSE3(const uint8* src, int src_stride, static void TransposeWx8_FAST_SSSE3(const uint8* src, int src_stride,
uint8* dst, int dst_stride, int width) { uint8* dst, int dst_stride, int width) {
asm volatile ( asm volatile(
// Read in the data from the source pointer. // Read in the data from the source pointer.
// First round of bit swap. // First round of bit swap.
"1: \n" "1: \n"
@ -646,7 +646,7 @@ static void TransposeUVWx8_SSE2(const uint8* src, int src_stride,
uint8* dst_a, int dst_stride_a, uint8* dst_a, int dst_stride_a,
uint8* dst_b, int dst_stride_b, uint8* dst_b, int dst_stride_b,
int w) { int w) {
asm volatile ( asm volatile(
// Read in the data from the source pointer. // Read in the data from the source pointer.
// First round of bit swap. // First round of bit swap.
"1: \n" "1: \n"

6
source/rotate_neon.cc
View File

@ -8,7 +8,7 @@
* be found in the AUTHORS file in the root of the source tree. * be found in the AUTHORS file in the root of the source tree.
*/ */
#include "row.h" #include "source/row.h"
#include "libyuv/basic_types.h" #include "libyuv/basic_types.h"
@ -25,7 +25,7 @@ static const uvec8 vtbl_4x4_transpose =
void TransposeWx8_NEON(const uint8* src, int src_stride, void TransposeWx8_NEON(const uint8* src, int src_stride,
uint8* dst, int dst_stride, uint8* dst, int dst_stride,
int width) { int width) {
asm volatile ( asm volatile(
// loops are on blocks of 8. loop will stop when // loops are on blocks of 8. loop will stop when
// counter gets to or below 0. starting the counter // counter gets to or below 0. starting the counter
// at w-8 allow for this // at w-8 allow for this
@ -191,7 +191,7 @@ void TransposeUVWx8_NEON(const uint8* src, int src_stride,
uint8* dst_a, int dst_stride_a, uint8* dst_a, int dst_stride_a,
uint8* dst_b, int dst_stride_b, uint8* dst_b, int dst_stride_b,
int width) { int width) {
asm volatile ( asm volatile(
// loops are on blocks of 8. loop will stop when // loops are on blocks of 8. loop will stop when
// counter gets to or below 0. starting the counter // counter gets to or below 0. starting the counter
// at w-8 allow for this // at w-8 allow for this

2
source/row_common.cc
View File

@ -8,7 +8,7 @@
* be found in the AUTHORS file in the root of the source tree. * be found in the AUTHORS file in the root of the source tree.
*/ */
#include "row.h" #include "source/row.h"
#include "libyuv/basic_types.h" #include "libyuv/basic_types.h"
#include <string.h> // For memcpy #include <string.h> // For memcpy

16
source/row_neon.cc
View File

@ -8,7 +8,7 @@
* be found in the AUTHORS file in the root of the source tree. * be found in the AUTHORS file in the root of the source tree.
*/ */
#include "row.h" #include "source/row.h"
#ifdef __cplusplus #ifdef __cplusplus
namespace libyuv { namespace libyuv {
@ -61,7 +61,7 @@ void I420ToARGBRow_NEON(const uint8* y_buf,
const uint8* v_buf, const uint8* v_buf,
uint8* rgb_buf, uint8* rgb_buf,
int width) { int width) {
asm volatile ( asm volatile(
"vld1.u8 {d24}, [%5] \n" "vld1.u8 {d24}, [%5] \n"
"vld1.u8 {d25}, [%6] \n" "vld1.u8 {d25}, [%6] \n"
"vmov.u8 d26, #128 \n" "vmov.u8 d26, #128 \n"
@ -93,7 +93,7 @@ void I420ToBGRARow_NEON(const uint8* y_buf,
const uint8* v_buf, const uint8* v_buf,
uint8* rgb_buf, uint8* rgb_buf,
int width) { int width) {
asm volatile ( asm volatile(
"vld1.u8 {d24}, [%5] \n" "vld1.u8 {d24}, [%5] \n"
"vld1.u8 {d25}, [%6] \n" "vld1.u8 {d25}, [%6] \n"
"vmov.u8 d26, #128 \n" "vmov.u8 d26, #128 \n"
@ -126,7 +126,7 @@ void I420ToABGRRow_NEON(const uint8* y_buf,
const uint8* v_buf, const uint8* v_buf,
uint8* rgb_buf, uint8* rgb_buf,
int width) { int width) {
asm volatile ( asm volatile(
"vld1.u8 {d24}, [%5] \n" "vld1.u8 {d24}, [%5] \n"
"vld1.u8 {d25}, [%6] \n" "vld1.u8 {d25}, [%6] \n"
"vmov.u8 d26, #128 \n" "vmov.u8 d26, #128 \n"
@ -157,7 +157,7 @@ YUVTORGB
// Reads 16 pairs of UV and write even values to dst_u and odd to dst_v // Reads 16 pairs of UV and write even values to dst_u and odd to dst_v
// Alignment requirement: 16 bytes for pointers, and multiple of 16 pixels. // Alignment requirement: 16 bytes for pointers, and multiple of 16 pixels.
void SplitUV_NEON(const uint8* src_uv, uint8* dst_u, uint8* dst_v, int width) { void SplitUV_NEON(const uint8* src_uv, uint8* dst_u, uint8* dst_v, int width) {
asm volatile ( asm volatile(
"1: \n" "1: \n"
"vld2.u8 {q0,q1}, [%0]! \n" // load 16 pairs of UV "vld2.u8 {q0,q1}, [%0]! \n" // load 16 pairs of UV
"subs %3, %3, #16 \n" // 16 processed per loop "subs %3, %3, #16 \n" // 16 processed per loop
@ -177,7 +177,7 @@ void SplitUV_NEON(const uint8* src_uv, uint8* dst_u, uint8* dst_v, int width) {
#ifdef HAS_COPYROW_NEON #ifdef HAS_COPYROW_NEON
// Copy multiple of 64 // Copy multiple of 64
void CopyRow_NEON(const uint8* src, uint8* dst, int count) { void CopyRow_NEON(const uint8* src, uint8* dst, int count) {
asm volatile ( asm volatile(
"1: \n" "1: \n"
"pld [%0, #0xC0] \n" // preload "pld [%0, #0xC0] \n" // preload
"vldm %0!,{q0,q1,q2,q3} \n" // load 64 "vldm %0!,{q0,q1,q2,q3} \n" // load 64
@ -195,7 +195,7 @@ void CopyRow_NEON(const uint8* src, uint8* dst, int count) {
#ifdef HAS_MIRRORROW_NEON #ifdef HAS_MIRRORROW_NEON
void MirrorRow_NEON(const uint8* src, uint8* dst, int width) { void MirrorRow_NEON(const uint8* src, uint8* dst, int width) {
asm volatile ( asm volatile(
// compute where to start writing destination // compute where to start writing destination
"add %1, %2 \n" "add %1, %2 \n"
// work on segments that are multiples of 16 // work on segments that are multiples of 16
@ -270,7 +270,7 @@ void MirrorRow_NEON(const uint8* src, uint8* dst, int width) {
#ifdef HAS_MIRRORROWUV_NEON #ifdef HAS_MIRRORROWUV_NEON
void MirrorRowUV_NEON(const uint8* src, uint8* dst_a, uint8* dst_b, int width) { void MirrorRowUV_NEON(const uint8* src, uint8* dst_a, uint8* dst_b, int width) {
asm volatile ( asm volatile(
// compute where to start writing destination // compute where to start writing destination
"add %1, %3 \n" // dst_a + width "add %1, %3 \n" // dst_a + width
"add %2, %3 \n" // dst_b + width "add %2, %3 \n" // dst_b + width

106
source/row_posix.cc
View File

@ -8,7 +8,7 @@
* be found in the AUTHORS file in the root of the source tree. * be found in the AUTHORS file in the root of the source tree.
*/ */
#include "row.h" #include "source/row.h"
#include "libyuv/basic_types.h" #include "libyuv/basic_types.h"
@ -109,7 +109,7 @@ CONST uvec8 kShuffleMaskARGBToRAW = {
}; };
void I400ToARGBRow_SSE2(const uint8* src_y, uint8* dst_argb, int pix) { void I400ToARGBRow_SSE2(const uint8* src_y, uint8* dst_argb, int pix) {
asm volatile ( asm volatile(
"pcmpeqb %%xmm5,%%xmm5 \n" "pcmpeqb %%xmm5,%%xmm5 \n"
"pslld $0x18,%%xmm5 \n" "pslld $0x18,%%xmm5 \n"
"1: \n" "1: \n"
@ -138,7 +138,7 @@ void I400ToARGBRow_SSE2(const uint8* src_y, uint8* dst_argb, int pix) {
} }
void ABGRToARGBRow_SSSE3(const uint8* src_abgr, uint8* dst_argb, int pix) { void ABGRToARGBRow_SSSE3(const uint8* src_abgr, uint8* dst_argb, int pix) {
asm volatile ( asm volatile(
"movdqa %3,%%xmm5 \n" "movdqa %3,%%xmm5 \n"
"sub %0,%1 \n" "sub %0,%1 \n"
"1: \n" "1: \n"
@ -161,7 +161,7 @@ void ABGRToARGBRow_SSSE3(const uint8* src_abgr, uint8* dst_argb, int pix) {
} }
void BGRAToARGBRow_SSSE3(const uint8* src_bgra, uint8* dst_argb, int pix) { void BGRAToARGBRow_SSSE3(const uint8* src_bgra, uint8* dst_argb, int pix) {
asm volatile ( asm volatile(
"movdqa %3,%%xmm5 \n" "movdqa %3,%%xmm5 \n"
"sub %0,%1 \n" "sub %0,%1 \n"
"1: \n" "1: \n"
@ -183,7 +183,7 @@ void BGRAToARGBRow_SSSE3(const uint8* src_bgra, uint8* dst_argb, int pix) {
} }
void RGB24ToARGBRow_SSSE3(const uint8* src_rgb24, uint8* dst_argb, int pix) { void RGB24ToARGBRow_SSSE3(const uint8* src_rgb24, uint8* dst_argb, int pix) {
asm volatile ( asm volatile(
"pcmpeqb %%xmm5,%%xmm5 \n" // generate mask 0xff000000 "pcmpeqb %%xmm5,%%xmm5 \n" // generate mask 0xff000000
"pslld $0x18,%%xmm5 \n" "pslld $0x18,%%xmm5 \n"
"movdqa %3,%%xmm4 \n" "movdqa %3,%%xmm4 \n"
@ -223,7 +223,7 @@ void RGB24ToARGBRow_SSSE3(const uint8* src_rgb24, uint8* dst_argb, int pix) {
} }
void RAWToARGBRow_SSSE3(const uint8* src_raw, uint8* dst_argb, int pix) { void RAWToARGBRow_SSSE3(const uint8* src_raw, uint8* dst_argb, int pix) {
asm volatile ( asm volatile(
"pcmpeqb %%xmm5,%%xmm5 \n" // generate mask 0xff000000 "pcmpeqb %%xmm5,%%xmm5 \n" // generate mask 0xff000000
"pslld $0x18,%%xmm5 \n" "pslld $0x18,%%xmm5 \n"
"movdqa %3,%%xmm4 \n" "movdqa %3,%%xmm4 \n"
@ -263,7 +263,7 @@ void RAWToARGBRow_SSSE3(const uint8* src_raw, uint8* dst_argb, int pix) {
} }
void RGB565ToARGBRow_SSE2(const uint8* src, uint8* dst, int pix) { void RGB565ToARGBRow_SSE2(const uint8* src, uint8* dst, int pix) {
asm volatile ( asm volatile(
"mov $0x1080108,%%eax \n" "mov $0x1080108,%%eax \n"
"movd %%eax,%%xmm5 \n" "movd %%eax,%%xmm5 \n"
"pshufd $0x0,%%xmm5,%%xmm5 \n" "pshufd $0x0,%%xmm5,%%xmm5 \n"
@ -312,7 +312,7 @@ void RGB565ToARGBRow_SSE2(const uint8* src, uint8* dst, int pix) {
} }
void ARGB1555ToARGBRow_SSE2(const uint8* src, uint8* dst, int pix) { void ARGB1555ToARGBRow_SSE2(const uint8* src, uint8* dst, int pix) {
asm volatile ( asm volatile(
"mov $0x1080108,%%eax \n" "mov $0x1080108,%%eax \n"
"movd %%eax,%%xmm5 \n" "movd %%eax,%%xmm5 \n"
"pshufd $0x0,%%xmm5,%%xmm5 \n" "pshufd $0x0,%%xmm5,%%xmm5 \n"
@ -364,7 +364,7 @@ void ARGB1555ToARGBRow_SSE2(const uint8* src, uint8* dst, int pix) {
} }
void ARGB4444ToARGBRow_SSE2(const uint8* src, uint8* dst, int pix) { void ARGB4444ToARGBRow_SSE2(const uint8* src, uint8* dst, int pix) {
asm volatile ( asm volatile(
"mov $0xf0f0f0f,%%eax \n" "mov $0xf0f0f0f,%%eax \n"
"movd %%eax,%%xmm4 \n" "movd %%eax,%%xmm4 \n"
"pshufd $0x0,%%xmm4,%%xmm4 \n" "pshufd $0x0,%%xmm4,%%xmm4 \n"
@ -403,7 +403,7 @@ void ARGB4444ToARGBRow_SSE2(const uint8* src, uint8* dst, int pix) {
} }
void ARGBToRGB24Row_SSSE3(const uint8* src, uint8* dst, int pix) { void ARGBToRGB24Row_SSSE3(const uint8* src, uint8* dst, int pix) {
asm volatile ( asm volatile(
"movdqa %3,%%xmm6 \n" "movdqa %3,%%xmm6 \n"
"1: \n" "1: \n"
"movdqa (%0),%%xmm0 \n" "movdqa (%0),%%xmm0 \n"
@ -443,7 +443,7 @@ void ARGBToRGB24Row_SSSE3(const uint8* src, uint8* dst, int pix) {
} }
void ARGBToRAWRow_SSSE3(const uint8* src, uint8* dst, int pix) { void ARGBToRAWRow_SSSE3(const uint8* src, uint8* dst, int pix) {
asm volatile ( asm volatile(
"movdqa %3,%%xmm6 \n" "movdqa %3,%%xmm6 \n"
"1: \n" "1: \n"
"movdqa (%0),%%xmm0 \n" "movdqa (%0),%%xmm0 \n"
@ -483,7 +483,7 @@ void ARGBToRAWRow_SSSE3(const uint8* src, uint8* dst, int pix) {
} }
void ARGBToRGB565Row_SSE2(const uint8* src, uint8* dst, int pix) { void ARGBToRGB565Row_SSE2(const uint8* src, uint8* dst, int pix) {
asm volatile ( asm volatile(
"pcmpeqb %%xmm3,%%xmm3 \n" "pcmpeqb %%xmm3,%%xmm3 \n"
"psrld $0x1b,%%xmm3 \n" "psrld $0x1b,%%xmm3 \n"
"pcmpeqb %%xmm4,%%xmm4 \n" "pcmpeqb %%xmm4,%%xmm4 \n"
@ -522,7 +522,7 @@ void ARGBToRGB565Row_SSE2(const uint8* src, uint8* dst, int pix) {
} }
void ARGBToARGB1555Row_SSE2(const uint8* src, uint8* dst, int pix) { void ARGBToARGB1555Row_SSE2(const uint8* src, uint8* dst, int pix) {
asm volatile ( asm volatile(
"pcmpeqb %%xmm4,%%xmm4 \n" "pcmpeqb %%xmm4,%%xmm4 \n"
"psrld $0x1b,%%xmm4 \n" "psrld $0x1b,%%xmm4 \n"
"movdqa %%xmm4,%%xmm5 \n" "movdqa %%xmm4,%%xmm5 \n"
@ -565,7 +565,7 @@ void ARGBToARGB1555Row_SSE2(const uint8* src, uint8* dst, int pix) {
} }
void ARGBToARGB4444Row_SSE2(const uint8* src, uint8* dst, int pix) { void ARGBToARGB4444Row_SSE2(const uint8* src, uint8* dst, int pix) {
asm volatile ( asm volatile(
"pcmpeqb %%xmm4,%%xmm4 \n" "pcmpeqb %%xmm4,%%xmm4 \n"
"psllw $0xc,%%xmm4 \n" "psllw $0xc,%%xmm4 \n"
"movdqa %%xmm4,%%xmm3 \n" "movdqa %%xmm4,%%xmm3 \n"
@ -596,7 +596,7 @@ void ARGBToARGB4444Row_SSE2(const uint8* src, uint8* dst, int pix) {
} }
void ARGBToYRow_SSSE3(const uint8* src_argb, uint8* dst_y, int pix) { void ARGBToYRow_SSSE3(const uint8* src_argb, uint8* dst_y, int pix) {
asm volatile ( asm volatile(
"movdqa %4,%%xmm5 \n" "movdqa %4,%%xmm5 \n"
"movdqa %3,%%xmm4 \n" "movdqa %3,%%xmm4 \n"
"1: \n" "1: \n"
@ -632,7 +632,7 @@ void ARGBToYRow_SSSE3(const uint8* src_argb, uint8* dst_y, int pix) {
} }
void ARGBToYRow_Unaligned_SSSE3(const uint8* src_argb, uint8* dst_y, int pix) { void ARGBToYRow_Unaligned_SSSE3(const uint8* src_argb, uint8* dst_y, int pix) {
asm volatile ( asm volatile(
"movdqa %4,%%xmm5 \n" "movdqa %4,%%xmm5 \n"
"movdqa %3,%%xmm4 \n" "movdqa %3,%%xmm4 \n"
"1: \n" "1: \n"
@ -674,7 +674,7 @@ void ARGBToYRow_Unaligned_SSSE3(const uint8* src_argb, uint8* dst_y, int pix) {
// and considered unsafe. // and considered unsafe.
void ARGBToUVRow_SSSE3(const uint8* src_argb0, int src_stride_argb, void ARGBToUVRow_SSSE3(const uint8* src_argb0, int src_stride_argb,
uint8* dst_u, uint8* dst_v, int width) { uint8* dst_u, uint8* dst_v, int width) {
asm volatile ( asm volatile(
"movdqa %0,%%xmm4 \n" "movdqa %0,%%xmm4 \n"
"movdqa %1,%%xmm3 \n" "movdqa %1,%%xmm3 \n"
"movdqa %2,%%xmm5 \n" "movdqa %2,%%xmm5 \n"
@ -687,7 +687,7 @@ void ARGBToUVRow_SSSE3(const uint8* src_argb0, int src_stride_argb,
"xmm3", "xmm4", "xmm5" "xmm3", "xmm4", "xmm5"
#endif #endif
); );
asm volatile ( asm volatile(
"sub %1,%2 \n" "sub %1,%2 \n"
"1: \n" "1: \n"
"movdqa (%0),%%xmm0 \n" "movdqa (%0),%%xmm0 \n"
@ -738,7 +738,7 @@ void ARGBToUVRow_SSSE3(const uint8* src_argb0, int src_stride_argb,
void ARGBToUVRow_Unaligned_SSSE3(const uint8* src_argb0, int src_stride_argb, void ARGBToUVRow_Unaligned_SSSE3(const uint8* src_argb0, int src_stride_argb,
uint8* dst_u, uint8* dst_v, int width) { uint8* dst_u, uint8* dst_v, int width) {
asm volatile ( asm volatile(
"movdqa %0,%%xmm4 \n" "movdqa %0,%%xmm4 \n"
"movdqa %1,%%xmm3 \n" "movdqa %1,%%xmm3 \n"
"movdqa %2,%%xmm5 \n" "movdqa %2,%%xmm5 \n"
@ -751,7 +751,7 @@ void ARGBToUVRow_Unaligned_SSSE3(const uint8* src_argb0, int src_stride_argb,
"xmm3", "xmm4", "xmm5" "xmm3", "xmm4", "xmm5"
#endif #endif
); );
asm volatile ( asm volatile(
"sub %1,%2 \n" "sub %1,%2 \n"
"1: \n" "1: \n"
"movdqu (%0),%%xmm0 \n" "movdqu (%0),%%xmm0 \n"
@ -805,7 +805,7 @@ void ARGBToUVRow_Unaligned_SSSE3(const uint8* src_argb0, int src_stride_argb,
} }
void BGRAToYRow_SSSE3(const uint8* src_bgra, uint8* dst_y, int pix) { void BGRAToYRow_SSSE3(const uint8* src_bgra, uint8* dst_y, int pix) {
asm volatile ( asm volatile(
"movdqa %4,%%xmm5 \n" "movdqa %4,%%xmm5 \n"
"movdqa %3,%%xmm4 \n" "movdqa %3,%%xmm4 \n"
"1: \n" "1: \n"
@ -841,7 +841,7 @@ void BGRAToYRow_SSSE3(const uint8* src_bgra, uint8* dst_y, int pix) {
} }
void BGRAToYRow_Unaligned_SSSE3(const uint8* src_bgra, uint8* dst_y, int pix) { void BGRAToYRow_Unaligned_SSSE3(const uint8* src_bgra, uint8* dst_y, int pix) {
asm volatile ( asm volatile(
"movdqa %4,%%xmm5 \n" "movdqa %4,%%xmm5 \n"
"movdqa %3,%%xmm4 \n" "movdqa %3,%%xmm4 \n"
"1: \n" "1: \n"
@ -878,7 +878,7 @@ void BGRAToYRow_Unaligned_SSSE3(const uint8* src_bgra, uint8* dst_y, int pix) {
void BGRAToUVRow_SSSE3(const uint8* src_bgra0, int src_stride_bgra, void BGRAToUVRow_SSSE3(const uint8* src_bgra0, int src_stride_bgra,
uint8* dst_u, uint8* dst_v, int width) { uint8* dst_u, uint8* dst_v, int width) {
asm volatile ( asm volatile(
"movdqa %0,%%xmm4 \n" "movdqa %0,%%xmm4 \n"
"movdqa %1,%%xmm3 \n" "movdqa %1,%%xmm3 \n"
"movdqa %2,%%xmm5 \n" "movdqa %2,%%xmm5 \n"
@ -891,7 +891,7 @@ void BGRAToUVRow_SSSE3(const uint8* src_bgra0, int src_stride_bgra,
"xmm3", "xmm4", "xmm5" "xmm3", "xmm4", "xmm5"
#endif #endif
); );
asm volatile ( asm volatile(
"sub %1,%2 \n" "sub %1,%2 \n"
"1: \n" "1: \n"
"movdqa (%0),%%xmm0 \n" "movdqa (%0),%%xmm0 \n"
@ -942,7 +942,7 @@ void BGRAToUVRow_SSSE3(const uint8* src_bgra0, int src_stride_bgra,
void BGRAToUVRow_Unaligned_SSSE3(const uint8* src_bgra0, int src_stride_bgra, void BGRAToUVRow_Unaligned_SSSE3(const uint8* src_bgra0, int src_stride_bgra,
uint8* dst_u, uint8* dst_v, int width) { uint8* dst_u, uint8* dst_v, int width) {
asm volatile ( asm volatile(
"movdqa %0,%%xmm4 \n" "movdqa %0,%%xmm4 \n"
"movdqa %1,%%xmm3 \n" "movdqa %1,%%xmm3 \n"
"movdqa %2,%%xmm5 \n" "movdqa %2,%%xmm5 \n"
@ -955,7 +955,7 @@ void BGRAToUVRow_Unaligned_SSSE3(const uint8* src_bgra0, int src_stride_bgra,
"xmm3", "xmm4", "xmm5" "xmm3", "xmm4", "xmm5"
#endif #endif
); );
asm volatile ( asm volatile(
"sub %1,%2 \n" "sub %1,%2 \n"
"1: \n" "1: \n"
"movdqu (%0),%%xmm0 \n" "movdqu (%0),%%xmm0 \n"
@ -1009,7 +1009,7 @@ void BGRAToUVRow_Unaligned_SSSE3(const uint8* src_bgra0, int src_stride_bgra,
} }
void ABGRToYRow_SSSE3(const uint8* src_abgr, uint8* dst_y, int pix) { void ABGRToYRow_SSSE3(const uint8* src_abgr, uint8* dst_y, int pix) {
asm volatile ( asm volatile(
"movdqa %4,%%xmm5 \n" "movdqa %4,%%xmm5 \n"
"movdqa %3,%%xmm4 \n" "movdqa %3,%%xmm4 \n"
"1: \n" "1: \n"
@ -1045,7 +1045,7 @@ void ABGRToYRow_SSSE3(const uint8* src_abgr, uint8* dst_y, int pix) {
} }
void ABGRToYRow_Unaligned_SSSE3(const uint8* src_abgr, uint8* dst_y, int pix) { void ABGRToYRow_Unaligned_SSSE3(const uint8* src_abgr, uint8* dst_y, int pix) {
asm volatile ( asm volatile(
"movdqa %4,%%xmm5 \n" "movdqa %4,%%xmm5 \n"
"movdqa %3,%%xmm4 \n" "movdqa %3,%%xmm4 \n"
"1: \n" "1: \n"
@ -1082,7 +1082,7 @@ void ABGRToYRow_Unaligned_SSSE3(const uint8* src_abgr, uint8* dst_y, int pix) {
void ABGRToUVRow_SSSE3(const uint8* src_abgr0, int src_stride_abgr, void ABGRToUVRow_SSSE3(const uint8* src_abgr0, int src_stride_abgr,
uint8* dst_u, uint8* dst_v, int width) { uint8* dst_u, uint8* dst_v, int width) {
asm volatile ( asm volatile(
"movdqa %0,%%xmm4 \n" "movdqa %0,%%xmm4 \n"
"movdqa %1,%%xmm3 \n" "movdqa %1,%%xmm3 \n"
"movdqa %2,%%xmm5 \n" "movdqa %2,%%xmm5 \n"
@ -1095,7 +1095,7 @@ void ABGRToUVRow_SSSE3(const uint8* src_abgr0, int src_stride_abgr,
"xmm3", "xmm4", "xmm5" "xmm3", "xmm4", "xmm5"
#endif #endif
); );
asm volatile ( asm volatile(
"sub %1,%2 \n" "sub %1,%2 \n"
"1: \n" "1: \n"
"movdqa (%0),%%xmm0 \n" "movdqa (%0),%%xmm0 \n"
@ -1146,7 +1146,7 @@ void ABGRToUVRow_SSSE3(const uint8* src_abgr0, int src_stride_abgr,
void ABGRToUVRow_Unaligned_SSSE3(const uint8* src_abgr0, int src_stride_abgr, void ABGRToUVRow_Unaligned_SSSE3(const uint8* src_abgr0, int src_stride_abgr,
uint8* dst_u, uint8* dst_v, int width) { uint8* dst_u, uint8* dst_v, int width) {
asm volatile ( asm volatile(
"movdqa %0,%%xmm4 \n" "movdqa %0,%%xmm4 \n"
"movdqa %1,%%xmm3 \n" "movdqa %1,%%xmm3 \n"
"movdqa %2,%%xmm5 \n" "movdqa %2,%%xmm5 \n"
@ -1159,7 +1159,7 @@ void ABGRToUVRow_Unaligned_SSSE3(const uint8* src_abgr0, int src_stride_abgr,
"xmm3", "xmm4", "xmm5" "xmm3", "xmm4", "xmm5"
#endif #endif
); );
asm volatile ( asm volatile(
"sub %1,%2 \n" "sub %1,%2 \n"
"1: \n" "1: \n"
"movdqu (%0),%%xmm0 \n" "movdqu (%0),%%xmm0 \n"
@ -1291,7 +1291,7 @@ void OMITFP I420ToARGBRow_SSSE3(const uint8* y_buf,
const uint8* v_buf, const uint8* v_buf,
uint8* rgb_buf, uint8* rgb_buf,
int width) { int width) {
asm volatile ( asm volatile(
"sub %1,%2 \n" "sub %1,%2 \n"
"pcmpeqb %%xmm5,%%xmm5 \n" "pcmpeqb %%xmm5,%%xmm5 \n"
"pxor %%xmm4,%%xmm4 \n" "pxor %%xmm4,%%xmm4 \n"
@ -1325,7 +1325,7 @@ void OMITFP I420ToBGRARow_SSSE3(const uint8* y_buf,
const uint8* v_buf, const uint8* v_buf,
uint8* rgb_buf, uint8* rgb_buf,
int width) { int width) {
asm volatile ( asm volatile(
"sub %1,%2 \n" "sub %1,%2 \n"
"pcmpeqb %%xmm5,%%xmm5 \n" "pcmpeqb %%xmm5,%%xmm5 \n"
"pxor %%xmm4,%%xmm4 \n" "pxor %%xmm4,%%xmm4 \n"
@ -1360,7 +1360,7 @@ void OMITFP I420ToABGRRow_SSSE3(const uint8* y_buf,
const uint8* v_buf, const uint8* v_buf,
uint8* rgb_buf, uint8* rgb_buf,
int width) { int width) {
asm volatile ( asm volatile(
"sub %1,%2 \n" "sub %1,%2 \n"
"pcmpeqb %%xmm5,%%xmm5 \n" "pcmpeqb %%xmm5,%%xmm5 \n"
"pxor %%xmm4,%%xmm4 \n" "pxor %%xmm4,%%xmm4 \n"
@ -1394,7 +1394,7 @@ void OMITFP I444ToARGBRow_SSSE3(const uint8* y_buf,
const uint8* v_buf, const uint8* v_buf,
uint8* rgb_buf, uint8* rgb_buf,
int width) { int width) {
asm volatile ( asm volatile(
"sub %1,%2 \n" "sub %1,%2 \n"
"pcmpeqb %%xmm5,%%xmm5 \n" "pcmpeqb %%xmm5,%%xmm5 \n"
"pxor %%xmm4,%%xmm4 \n" "pxor %%xmm4,%%xmm4 \n"
@ -1450,7 +1450,7 @@ void OMITFP I444ToARGBRow_SSSE3(const uint8* y_buf,
void YToARGBRow_SSE2(const uint8* y_buf, void YToARGBRow_SSE2(const uint8* y_buf,
uint8* rgb_buf, uint8* rgb_buf,
int width) { int width) {
asm volatile ( asm volatile(
"pcmpeqb %%xmm4,%%xmm4 \n" "pcmpeqb %%xmm4,%%xmm4 \n"
"pslld $0x18,%%xmm4 \n" "pslld $0x18,%%xmm4 \n"
"mov $0x10001000,%%eax \n" "mov $0x10001000,%%eax \n"
@ -1501,7 +1501,7 @@ CONST uvec8 kShuffleMirror = {
void MirrorRow_SSSE3(const uint8* src, uint8* dst, int width) { void MirrorRow_SSSE3(const uint8* src, uint8* dst, int width) {
intptr_t temp_width = static_cast<intptr_t>(width); intptr_t temp_width = static_cast<intptr_t>(width);
asm volatile ( asm volatile(
"movdqa %3,%%xmm5 \n" "movdqa %3,%%xmm5 \n"
"lea -0x10(%0),%0 \n" "lea -0x10(%0),%0 \n"
"1: \n" "1: \n"
@ -1526,7 +1526,7 @@ void MirrorRow_SSSE3(const uint8* src, uint8* dst, int width) {
#ifdef HAS_MIRRORROW_SSE2 #ifdef HAS_MIRRORROW_SSE2
void MirrorRow_SSE2(const uint8* src, uint8* dst, int width) { void MirrorRow_SSE2(const uint8* src, uint8* dst, int width) {
intptr_t temp_width = static_cast<intptr_t>(width); intptr_t temp_width = static_cast<intptr_t>(width);
asm volatile ( asm volatile(
"lea -0x10(%0),%0 \n" "lea -0x10(%0),%0 \n"
"1: \n" "1: \n"
"movdqu (%0,%2),%%xmm0 \n" "movdqu (%0,%2),%%xmm0 \n"
@ -1561,7 +1561,7 @@ CONST uvec8 kShuffleMirrorUV = {
void MirrorRowUV_SSSE3(const uint8* src, uint8* dst_u, uint8* dst_v, void MirrorRowUV_SSSE3(const uint8* src, uint8* dst_u, uint8* dst_v,
int width) { int width) {
intptr_t temp_width = static_cast<intptr_t>(width); intptr_t temp_width = static_cast<intptr_t>(width);
asm volatile ( asm volatile(
"movdqa %4,%%xmm1 \n" "movdqa %4,%%xmm1 \n"
"lea -16(%0,%3,2),%0 \n" "lea -16(%0,%3,2),%0 \n"
"sub %1,%2 \n" "sub %1,%2 \n"
@ -1589,7 +1589,7 @@ void MirrorRowUV_SSSE3(const uint8* src, uint8* dst_u, uint8* dst_v,
#ifdef HAS_SPLITUV_SSE2 #ifdef HAS_SPLITUV_SSE2
void SplitUV_SSE2(const uint8* src_uv, uint8* dst_u, uint8* dst_v, int pix) { void SplitUV_SSE2(const uint8* src_uv, uint8* dst_u, uint8* dst_v, int pix) {
asm volatile ( asm volatile(
"pcmpeqb %%xmm5,%%xmm5 \n" "pcmpeqb %%xmm5,%%xmm5 \n"
"psrlw $0x8,%%xmm5 \n" "psrlw $0x8,%%xmm5 \n"
"sub %1,%2 \n" "sub %1,%2 \n"
@ -1625,7 +1625,7 @@ void SplitUV_SSE2(const uint8* src_uv, uint8* dst_u, uint8* dst_v, int pix) {
#ifdef HAS_COPYROW_SSE2 #ifdef HAS_COPYROW_SSE2
void CopyRow_SSE2(const uint8* src, uint8* dst, int count) { void CopyRow_SSE2(const uint8* src, uint8* dst, int count) {
asm volatile ( asm volatile(
"sub %0,%1 \n" "sub %0,%1 \n"
"1: \n" "1: \n"
"movdqa (%0),%%xmm0 \n" "movdqa (%0),%%xmm0 \n"
@ -1650,7 +1650,7 @@ void CopyRow_SSE2(const uint8* src, uint8* dst, int count) {
#ifdef HAS_COPYROW_X86 #ifdef HAS_COPYROW_X86
void CopyRow_X86(const uint8* src, uint8* dst, int width) { void CopyRow_X86(const uint8* src, uint8* dst, int width) {
size_t width_tmp = static_cast<size_t>(width); size_t width_tmp = static_cast<size_t>(width);
asm volatile ( asm volatile(
"shr $0x2,%2 \n" "shr $0x2,%2 \n"
"rep movsl \n" "rep movsl \n"
: "+S"(src), // %0 : "+S"(src), // %0
@ -1664,7 +1664,7 @@ void CopyRow_X86(const uint8* src, uint8* dst, int width) {
#ifdef HAS_YUY2TOYROW_SSE2 #ifdef HAS_YUY2TOYROW_SSE2
void YUY2ToYRow_SSE2(const uint8* src_yuy2, uint8* dst_y, int pix) { void YUY2ToYRow_SSE2(const uint8* src_yuy2, uint8* dst_y, int pix) {
asm volatile ( asm volatile(
"pcmpeqb %%xmm5,%%xmm5 \n" "pcmpeqb %%xmm5,%%xmm5 \n"
"psrlw $0x8,%%xmm5 \n" "psrlw $0x8,%%xmm5 \n"
"1: \n" "1: \n"
@ -1691,7 +1691,7 @@ void YUY2ToYRow_SSE2(const uint8* src_yuy2, uint8* dst_y, int pix) {
void YUY2ToUVRow_SSE2(const uint8* src_yuy2, int stride_yuy2, void YUY2ToUVRow_SSE2(const uint8* src_yuy2, int stride_yuy2,
uint8* dst_u, uint8* dst_y, int pix) { uint8* dst_u, uint8* dst_y, int pix) {
asm volatile ( asm volatile(
"pcmpeqb %%xmm5,%%xmm5 \n" "pcmpeqb %%xmm5,%%xmm5 \n"
"psrlw $0x8,%%xmm5 \n" "psrlw $0x8,%%xmm5 \n"
"sub %1,%2 \n" "sub %1,%2 \n"
@ -1730,7 +1730,7 @@ void YUY2ToUVRow_SSE2(const uint8* src_yuy2, int stride_yuy2,
void YUY2ToYRow_Unaligned_SSE2(const uint8* src_yuy2, void YUY2ToYRow_Unaligned_SSE2(const uint8* src_yuy2,
uint8* dst_y, int pix) { uint8* dst_y, int pix) {
asm volatile ( asm volatile(
"pcmpeqb %%xmm5,%%xmm5 \n" "pcmpeqb %%xmm5,%%xmm5 \n"
"psrlw $0x8,%%xmm5 \n" "psrlw $0x8,%%xmm5 \n"
"1: \n" "1: \n"
@ -1759,7 +1759,7 @@ void YUY2ToUVRow_Unaligned_SSE2(const uint8* src_yuy2,
int stride_yuy2, int stride_yuy2,
uint8* dst_u, uint8* dst_y, uint8* dst_u, uint8* dst_y,
int pix) { int pix) {
asm volatile ( asm volatile(
"pcmpeqb %%xmm5,%%xmm5 \n" "pcmpeqb %%xmm5,%%xmm5 \n"
"psrlw $0x8,%%xmm5 \n" "psrlw $0x8,%%xmm5 \n"
"sub %1,%2 \n" "sub %1,%2 \n"
@ -1797,7 +1797,7 @@ void YUY2ToUVRow_Unaligned_SSE2(const uint8* src_yuy2,
} }
void UYVYToYRow_SSE2(const uint8* src_uyvy, uint8* dst_y, int pix) { void UYVYToYRow_SSE2(const uint8* src_uyvy, uint8* dst_y, int pix) {
asm volatile ( asm volatile(
"1: \n" "1: \n"
"movdqa (%0),%%xmm0 \n" "movdqa (%0),%%xmm0 \n"
"movdqa 0x10(%0),%%xmm1 \n" "movdqa 0x10(%0),%%xmm1 \n"
@ -1822,7 +1822,7 @@ void UYVYToYRow_SSE2(const uint8* src_uyvy, uint8* dst_y, int pix) {
void UYVYToUVRow_SSE2(const uint8* src_uyvy, int stride_uyvy, void UYVYToUVRow_SSE2(const uint8* src_uyvy, int stride_uyvy,
uint8* dst_u, uint8* dst_y, int pix) { uint8* dst_u, uint8* dst_y, int pix) {
asm volatile ( asm volatile(
"pcmpeqb %%xmm5,%%xmm5 \n" "pcmpeqb %%xmm5,%%xmm5 \n"
"psrlw $0x8,%%xmm5 \n" "psrlw $0x8,%%xmm5 \n"
"sub %1,%2 \n" "sub %1,%2 \n"
@ -1861,7 +1861,7 @@ void UYVYToUVRow_SSE2(const uint8* src_uyvy, int stride_uyvy,
void UYVYToYRow_Unaligned_SSE2(const uint8* src_uyvy, void UYVYToYRow_Unaligned_SSE2(const uint8* src_uyvy,
uint8* dst_y, int pix) { uint8* dst_y, int pix) {
asm volatile ( asm volatile(
"1: \n" "1: \n"
"movdqu (%0),%%xmm0 \n" "movdqu (%0),%%xmm0 \n"
"movdqu 0x10(%0),%%xmm1 \n" "movdqu 0x10(%0),%%xmm1 \n"
@ -1886,7 +1886,7 @@ void UYVYToYRow_Unaligned_SSE2(const uint8* src_uyvy,
void UYVYToUVRow_Unaligned_SSE2(const uint8* src_uyvy, int stride_uyvy, void UYVYToUVRow_Unaligned_SSE2(const uint8* src_uyvy, int stride_uyvy,
uint8* dst_u, uint8* dst_y, int pix) { uint8* dst_u, uint8* dst_y, int pix) {
asm volatile ( asm volatile(
"pcmpeqb %%xmm5,%%xmm5 \n" "pcmpeqb %%xmm5,%%xmm5 \n"
"psrlw $0x8,%%xmm5 \n" "psrlw $0x8,%%xmm5 \n"
"sub %1,%2 \n" "sub %1,%2 \n"
@ -1929,7 +1929,7 @@ void UYVYToUVRow_Unaligned_SSE2(const uint8* src_uyvy, int stride_uyvy,
// Destination aligned to 16 bytes, multiple of 4 pixels // Destination aligned to 16 bytes, multiple of 4 pixels
void ARGBBlendRow_Aligned_SSE2(const uint8* src_argb, uint8* dst_argb, void ARGBBlendRow_Aligned_SSE2(const uint8* src_argb, uint8* dst_argb,
int width) { int width) {
asm volatile ( asm volatile(
"pcmpeqb %%xmm7,%%xmm7 \n" "pcmpeqb %%xmm7,%%xmm7 \n"
"psrlw $0xf,%%xmm7 \n" "psrlw $0xf,%%xmm7 \n"
"pcmpeqb %%xmm6,%%xmm6 \n" "pcmpeqb %%xmm6,%%xmm6 \n"
@ -1999,7 +1999,7 @@ void ARGBBlendRow_Aligned_SSE2(const uint8* src_argb, uint8* dst_argb,
// Blend 1 pixel at a time, unaligned // Blend 1 pixel at a time, unaligned
void ARGBBlendRow1_SSE2(const uint8* src_argb, uint8* dst_argb, int width) { void ARGBBlendRow1_SSE2(const uint8* src_argb, uint8* dst_argb, int width) {
asm volatile ( asm volatile(
"pcmpeqb %%xmm7,%%xmm7 \n" "pcmpeqb %%xmm7,%%xmm7 \n"
"psrlw $0xf,%%xmm7 \n" "psrlw $0xf,%%xmm7 \n"
"pcmpeqb %%xmm6,%%xmm6 \n" "pcmpeqb %%xmm6,%%xmm6 \n"

2
source/row_win.cc
View File

@ -8,7 +8,7 @@
* be found in the AUTHORS file in the root of the source tree. * be found in the AUTHORS file in the root of the source tree.
*/ */
#include "row.h" #include "source/row.h"
#if defined(_M_IX86) #if defined(_M_IX86)
#include "emmintrin.h" #include "emmintrin.h"

62
source/scale.cc
View File

@ -16,7 +16,7 @@
#include "libyuv/cpu_id.h" #include "libyuv/cpu_id.h"
#include "libyuv/planar_functions.h" // For CopyPlane #include "libyuv/planar_functions.h" // For CopyPlane
#include "row.h" #include "source/row.h"
#ifdef __cplusplus #ifdef __cplusplus
namespace libyuv { namespace libyuv {
@ -59,7 +59,7 @@ void SetUseReferenceImpl(bool use) {
#define HAS_SCALEROWDOWN2_NEON #define HAS_SCALEROWDOWN2_NEON
void ScaleRowDown2_NEON(const uint8* src_ptr, int /* src_stride */, void ScaleRowDown2_NEON(const uint8* src_ptr, int /* src_stride */,
uint8* dst, int dst_width) { uint8* dst, int dst_width) {
asm volatile ( asm volatile(
"1: \n" "1: \n"
"vld2.u8 {q0,q1}, [%0]! \n" // load even pixels into q0, odd into q1 "vld2.u8 {q0,q1}, [%0]! \n" // load even pixels into q0, odd into q1
"vst1.u8 {q0}, [%1]! \n" // store even pixels "vst1.u8 {q0}, [%1]! \n" // store even pixels
@ -75,7 +75,7 @@ void ScaleRowDown2_NEON(const uint8* src_ptr, int /* src_stride */,
void ScaleRowDown2Int_NEON(const uint8* src_ptr, int src_stride, void ScaleRowDown2Int_NEON(const uint8* src_ptr, int src_stride,
uint8* dst, int dst_width) { uint8* dst, int dst_width) {
asm volatile ( asm volatile(
"add %1, %0 \n" // change the stride to row 2 pointer "add %1, %0 \n" // change the stride to row 2 pointer
"1: \n" "1: \n"
"vld1.u8 {q0,q1}, [%0]! \n" // load row 1 and post increment "vld1.u8 {q0,q1}, [%0]! \n" // load row 1 and post increment
@ -101,7 +101,7 @@ void ScaleRowDown2Int_NEON(const uint8* src_ptr, int src_stride,
#define HAS_SCALEROWDOWN4_NEON #define HAS_SCALEROWDOWN4_NEON
static void ScaleRowDown4_NEON(const uint8* src_ptr, int /* src_stride */, static void ScaleRowDown4_NEON(const uint8* src_ptr, int /* src_stride */,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile ( asm volatile(
"1: \n" "1: \n"
"vld2.u8 {d0, d1}, [%0]! \n" "vld2.u8 {d0, d1}, [%0]! \n"
"vtrn.u8 d1, d0 \n" "vtrn.u8 d1, d0 \n"
@ -120,7 +120,7 @@ static void ScaleRowDown4_NEON(const uint8* src_ptr, int /* src_stride */,
static void ScaleRowDown4Int_NEON(const uint8* src_ptr, int src_stride, static void ScaleRowDown4Int_NEON(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile ( asm volatile(
"add r4, %0, %3 \n" "add r4, %0, %3 \n"
"add r5, r4, %3 \n" "add r5, r4, %3 \n"
"add %3, r5, %3 \n" "add %3, r5, %3 \n"
@ -159,7 +159,7 @@ static void ScaleRowDown4Int_NEON(const uint8* src_ptr, int src_stride,
// Point samples 32 pixels to 24 pixels. // Point samples 32 pixels to 24 pixels.
static void ScaleRowDown34_NEON(const uint8* src_ptr, int /* src_stride */, static void ScaleRowDown34_NEON(const uint8* src_ptr, int /* src_stride */,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile ( asm volatile(
"1: \n" "1: \n"
"vld4.u8 {d0, d1, d2, d3}, [%0]! \n" // src line 0 "vld4.u8 {d0, d1, d2, d3}, [%0]! \n" // src line 0
"vmov d2, d3 \n" // order needs to be d0, d1, d2 "vmov d2, d3 \n" // order needs to be d0, d1, d2
@ -176,7 +176,7 @@ static void ScaleRowDown34_NEON(const uint8* src_ptr, int /* src_stride */,
static void ScaleRowDown34_0_Int_NEON(const uint8* src_ptr, int src_stride, static void ScaleRowDown34_0_Int_NEON(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile ( asm volatile(
"vmov.u8 d24, #3 \n" "vmov.u8 d24, #3 \n"
"add %3, %0 \n" "add %3, %0 \n"
"1: \n" "1: \n"
@ -231,7 +231,7 @@ static void ScaleRowDown34_0_Int_NEON(const uint8* src_ptr, int src_stride,
static void ScaleRowDown34_1_Int_NEON(const uint8* src_ptr, int src_stride, static void ScaleRowDown34_1_Int_NEON(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile ( asm volatile(
"vmov.u8 d24, #3 \n" "vmov.u8 d24, #3 \n"
"add %3, %0 \n" "add %3, %0 \n"
"1: \n" "1: \n"
@ -283,7 +283,7 @@ const unsigned short mult38_div9[8] __attribute__ ((aligned(16))) =
// 32 -> 12 // 32 -> 12
static void ScaleRowDown38_NEON(const uint8* src_ptr, int, static void ScaleRowDown38_NEON(const uint8* src_ptr, int,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile ( asm volatile(
"vld1.u8 {q3}, [%3] \n" "vld1.u8 {q3}, [%3] \n"
"1: \n" "1: \n"
"vld1.u8 {d0, d1, d2, d3}, [%0]! \n" "vld1.u8 {d0, d1, d2, d3}, [%0]! \n"
@ -304,7 +304,7 @@ static void ScaleRowDown38_NEON(const uint8* src_ptr, int,
// 32x3 -> 12x1 // 32x3 -> 12x1
static void ScaleRowDown38_3_Int_NEON(const uint8* src_ptr, int src_stride, static void ScaleRowDown38_3_Int_NEON(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile ( asm volatile(
"vld1.u16 {q13}, [%4] \n" "vld1.u16 {q13}, [%4] \n"
"vld1.u8 {q14}, [%5] \n" "vld1.u8 {q14}, [%5] \n"
"vld1.u8 {q15}, [%6] \n" "vld1.u8 {q15}, [%6] \n"
@ -413,7 +413,7 @@ static void ScaleRowDown38_3_Int_NEON(const uint8* src_ptr, int src_stride,
// 32x2 -> 12x1 // 32x2 -> 12x1
static void ScaleRowDown38_2_Int_NEON(const uint8* src_ptr, int src_stride, static void ScaleRowDown38_2_Int_NEON(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile ( asm volatile(
"vld1.u16 {q13}, [%4] \n" "vld1.u16 {q13}, [%4] \n"
"vld1.u8 {q14}, [%5] \n" "vld1.u8 {q14}, [%5] \n"
"add %3, %0 \n" "add %3, %0 \n"
@ -508,7 +508,7 @@ static void ScaleRowDown38_2_Int_NEON(const uint8* src_ptr, int src_stride,
static void ScaleFilterRows_NEON(uint8* dst_ptr, static void ScaleFilterRows_NEON(uint8* dst_ptr,
const uint8* src_ptr, int src_stride, const uint8* src_ptr, int src_stride,
int dst_width, int source_y_fraction) { int dst_width, int source_y_fraction) {
asm volatile ( asm volatile(
"cmp %4, #0 \n" "cmp %4, #0 \n"
"beq 2f \n" "beq 2f \n"
"add %2, %1 \n" "add %2, %1 \n"
@ -1555,7 +1555,7 @@ static void ScaleFilterCols34_SSSE3(uint8* dst_ptr, const uint8* src_ptr,
#define HAS_SCALEROWDOWN2_SSE2 #define HAS_SCALEROWDOWN2_SSE2
static void ScaleRowDown2_SSE2(const uint8* src_ptr, int src_stride, static void ScaleRowDown2_SSE2(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile ( asm volatile(
"pcmpeqb %%xmm5,%%xmm5 \n" "pcmpeqb %%xmm5,%%xmm5 \n"
"psrlw $0x8,%%xmm5 \n" "psrlw $0x8,%%xmm5 \n"
"1: \n" "1: \n"
@ -1579,7 +1579,7 @@ static void ScaleRowDown2_SSE2(const uint8* src_ptr, int src_stride,
static void ScaleRowDown2Int_SSE2(const uint8* src_ptr, int src_stride, static void ScaleRowDown2Int_SSE2(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile ( asm volatile(
"pcmpeqb %%xmm5,%%xmm5 \n" "pcmpeqb %%xmm5,%%xmm5 \n"
"psrlw $0x8,%%xmm5 \n" "psrlw $0x8,%%xmm5 \n"
"1: \n" "1: \n"
@ -1614,7 +1614,7 @@ static void ScaleRowDown2Int_SSE2(const uint8* src_ptr, int src_stride,
#define HAS_SCALEROWDOWN4_SSE2 #define HAS_SCALEROWDOWN4_SSE2
static void ScaleRowDown4_SSE2(const uint8* src_ptr, int src_stride, static void ScaleRowDown4_SSE2(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile ( asm volatile(
"pcmpeqb %%xmm5,%%xmm5 \n" "pcmpeqb %%xmm5,%%xmm5 \n"
"psrld $0x18,%%xmm5 \n" "psrld $0x18,%%xmm5 \n"
"1: \n" "1: \n"
@ -1640,7 +1640,7 @@ static void ScaleRowDown4_SSE2(const uint8* src_ptr, int src_stride,
static void ScaleRowDown4Int_SSE2(const uint8* src_ptr, int src_stride, static void ScaleRowDown4Int_SSE2(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
intptr_t temp = 0; intptr_t temp = 0;
asm volatile ( asm volatile(
"pcmpeqb %%xmm7,%%xmm7 \n" "pcmpeqb %%xmm7,%%xmm7 \n"
"psrlw $0x8,%%xmm7 \n" "psrlw $0x8,%%xmm7 \n"
"lea (%4,%4,2),%3 \n" "lea (%4,%4,2),%3 \n"
@ -1693,7 +1693,7 @@ static void ScaleRowDown4Int_SSE2(const uint8* src_ptr, int src_stride,
#define HAS_SCALEROWDOWN8_SSE2 #define HAS_SCALEROWDOWN8_SSE2
static void ScaleRowDown8_SSE2(const uint8* src_ptr, int src_stride, static void ScaleRowDown8_SSE2(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile ( asm volatile(
"pcmpeqb %%xmm5,%%xmm5 \n" "pcmpeqb %%xmm5,%%xmm5 \n"
"psrlq $0x38,%%xmm5 \n" "psrlq $0x38,%%xmm5 \n"
"1: \n" "1: \n"
@ -1722,7 +1722,7 @@ static void ScaleAddRows_SSE2(const uint8* src_ptr, int src_stride,
uint16* dst_ptr, int src_width, int src_height) { uint16* dst_ptr, int src_width, int src_height) {
int tmp_height = 0; int tmp_height = 0;
intptr_t tmp_src = 0; intptr_t tmp_src = 0;
asm volatile ( asm volatile(
"pxor %%xmm4,%%xmm4 \n" "pxor %%xmm4,%%xmm4 \n"
"sub $0x1,%5 \n" "sub $0x1,%5 \n"
"1: \n" "1: \n"
@ -2263,7 +2263,7 @@ extern "C" void ScaleFilterRows_SSSE3(uint8* dst_ptr,
#elif defined(__x86_64__) #elif defined(__x86_64__)
static void ScaleRowDown8Int_SSE2(const uint8* src_ptr, int src_stride, static void ScaleRowDown8Int_SSE2(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile ( asm volatile(
"lea (%3,%3,2),%%r10 \n" "lea (%3,%3,2),%%r10 \n"
"pxor %%xmm7,%%xmm7 \n" "pxor %%xmm7,%%xmm7 \n"
"1:" "1:"
@ -2322,7 +2322,7 @@ static void ScaleRowDown8Int_SSE2(const uint8* src_ptr, int src_stride,
#define HAS_SCALEROWDOWN34_SSSE3 #define HAS_SCALEROWDOWN34_SSSE3
static void ScaleRowDown34_SSSE3(const uint8* src_ptr, int src_stride, static void ScaleRowDown34_SSSE3(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile ( asm volatile(
"movdqa (%3),%%xmm3 \n" "movdqa (%3),%%xmm3 \n"
"movdqa (%4),%%xmm4 \n" "movdqa (%4),%%xmm4 \n"
"movdqa (%5),%%xmm5 \n" "movdqa (%5),%%xmm5 \n"
@ -2353,7 +2353,7 @@ static void ScaleRowDown34_SSSE3(const uint8* src_ptr, int src_stride,
static void ScaleRowDown34_1_Int_SSSE3(const uint8* src_ptr, int src_stride, static void ScaleRowDown34_1_Int_SSSE3(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile ( asm volatile(
"movdqa (%4),%%xmm2 \n" // _shuf01 "movdqa (%4),%%xmm2 \n" // _shuf01
"movdqa (%5),%%xmm3 \n" // _shuf11 "movdqa (%5),%%xmm3 \n" // _shuf11
"movdqa (%6),%%xmm4 \n" // _shuf21 "movdqa (%6),%%xmm4 \n" // _shuf21
@ -2410,7 +2410,7 @@ static void ScaleRowDown34_1_Int_SSSE3(const uint8* src_ptr, int src_stride,
static void ScaleRowDown34_0_Int_SSSE3(const uint8* src_ptr, int src_stride, static void ScaleRowDown34_0_Int_SSSE3(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile ( asm volatile(
"movdqa (%4),%%xmm2 \n" // _shuf01 "movdqa (%4),%%xmm2 \n" // _shuf01
"movdqa (%5),%%xmm3 \n" // _shuf11 "movdqa (%5),%%xmm3 \n" // _shuf11
"movdqa (%6),%%xmm4 \n" // _shuf21 "movdqa (%6),%%xmm4 \n" // _shuf21
@ -2471,7 +2471,7 @@ static void ScaleRowDown34_0_Int_SSSE3(const uint8* src_ptr, int src_stride,
#define HAS_SCALEROWDOWN38_SSSE3 #define HAS_SCALEROWDOWN38_SSSE3
static void ScaleRowDown38_SSSE3(const uint8* src_ptr, int src_stride, static void ScaleRowDown38_SSSE3(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile ( asm volatile(
"movdqa (%3),%%xmm4 \n" "movdqa (%3),%%xmm4 \n"
"movdqa (%4),%%xmm5 \n" "movdqa (%4),%%xmm5 \n"
"1:" "1:"
@ -2498,7 +2498,7 @@ static void ScaleRowDown38_SSSE3(const uint8* src_ptr, int src_stride,
static void ScaleRowDown38_3_Int_SSSE3(const uint8* src_ptr, int src_stride, static void ScaleRowDown38_3_Int_SSSE3(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile ( asm volatile(
"movdqa (%4),%%xmm4 \n" "movdqa (%4),%%xmm4 \n"
"movdqa (%5),%%xmm5 \n" "movdqa (%5),%%xmm5 \n"
"movdqa (%6),%%xmm6 \n" "movdqa (%6),%%xmm6 \n"
@ -2555,7 +2555,7 @@ static void ScaleRowDown38_3_Int_SSSE3(const uint8* src_ptr, int src_stride,
static void ScaleRowDown38_2_Int_SSSE3(const uint8* src_ptr, int src_stride, static void ScaleRowDown38_2_Int_SSSE3(const uint8* src_ptr, int src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile ( asm volatile(
"movdqa (%4),%%xmm4 \n" "movdqa (%4),%%xmm4 \n"
"movdqa (%5),%%xmm5 \n" "movdqa (%5),%%xmm5 \n"
"movdqa (%6),%%xmm6 \n" "movdqa (%6),%%xmm6 \n"
@ -2597,7 +2597,7 @@ static void ScaleFilterRows_SSE2(uint8* dst_ptr,
const uint8* src_ptr, int src_stride, const uint8* src_ptr, int src_stride,
int dst_width, int source_y_fraction) { int dst_width, int source_y_fraction) {
if (source_y_fraction == 0) { if (source_y_fraction == 0) {
asm volatile ( asm volatile(
"1:" "1:"
"movdqa (%1),%%xmm0 \n" "movdqa (%1),%%xmm0 \n"
"lea 0x10(%1),%1 \n" "lea 0x10(%1),%1 \n"
@ -2615,7 +2615,7 @@ static void ScaleFilterRows_SSE2(uint8* dst_ptr,
); );
return; return;
} else if (source_y_fraction == 128) { } else if (source_y_fraction == 128) {
asm volatile ( asm volatile(
"1:" "1:"
"movdqa (%1),%%xmm0 \n" "movdqa (%1),%%xmm0 \n"
"movdqa (%1,%3,1),%%xmm2 \n" "movdqa (%1,%3,1),%%xmm2 \n"
@ -2635,7 +2635,7 @@ static void ScaleFilterRows_SSE2(uint8* dst_ptr,
); );
return; return;
} else { } else {
asm volatile ( asm volatile(
"mov %3,%%eax \n" "mov %3,%%eax \n"
"movd %%eax,%%xmm6 \n" "movd %%eax,%%xmm6 \n"
"punpcklwd %%xmm6,%%xmm6 \n" "punpcklwd %%xmm6,%%xmm6 \n"
@ -2688,7 +2688,7 @@ static void ScaleFilterRows_SSSE3(uint8* dst_ptr,
const uint8* src_ptr, int src_stride, const uint8* src_ptr, int src_stride,
int dst_width, int source_y_fraction) { int dst_width, int source_y_fraction) {
if (source_y_fraction <= 1) { if (source_y_fraction <= 1) {
asm volatile ( asm volatile(
"1:" "1:"
"movdqa (%1),%%xmm0 \n" "movdqa (%1),%%xmm0 \n"
"lea 0x10(%1),%1 \n" "lea 0x10(%1),%1 \n"
@ -2706,7 +2706,7 @@ static void ScaleFilterRows_SSSE3(uint8* dst_ptr,
); );
return; return;
} else if (source_y_fraction == 128) { } else if (source_y_fraction == 128) {
asm volatile ( asm volatile(
"1:" "1:"
"movdqa (%1),%%xmm0 \n" "movdqa (%1),%%xmm0 \n"
"movdqa (%1,%3,1),%%xmm2 \n" "movdqa (%1,%3,1),%%xmm2 \n"
@ -2726,7 +2726,7 @@ static void ScaleFilterRows_SSSE3(uint8* dst_ptr,
); );
return; return;
} else { } else {
asm volatile ( asm volatile(
"mov %3,%%eax \n" "mov %3,%%eax \n"
"shr %%eax \n" "shr %%eax \n"
"mov %%al,%%ah \n" "mov %%al,%%ah \n"

11
unit_test/compare_test.cc
View File

@ -8,12 +8,11 @@
* be found in the AUTHORS file in the root of the source tree. * be found in the AUTHORS file in the root of the source tree.
*/ */
#include "unit_test.h"
#include <stdlib.h> #include <stdlib.h>
#include <string.h> #include <string.h>
#include <time.h> #include <time.h>
#include "unit_test/unit_test.h"
#include "libyuv/basic_types.h" #include "libyuv/basic_types.h"
#include "libyuv/compare.h" #include "libyuv/compare.h"
#include "libyuv/cpu_id.h" #include "libyuv/cpu_id.h"
@ -153,7 +152,7 @@ TEST_F(libyuvTest, BenchmarkPsnr_C) {
_benchmark_width, _benchmark_height); _benchmark_width, _benchmark_height);
c_time = (get_time() - c_time) / _benchmark_iterations; c_time = (get_time() - c_time) / _benchmark_iterations;
printf ("BenchmarkPsnr_C - %8d us c\n", (int)(c_time*1e6)); printf("BenchmarkPsnr_C - %8.2f us c\n", c_time * 1e6);
MaskCpuFlags(-1); MaskCpuFlags(-1);
@ -176,7 +175,7 @@ TEST_F(libyuvTest, BenchmarkPsnr_OPT) {
_benchmark_width, _benchmark_height); _benchmark_width, _benchmark_height);
opt_time = (get_time() - opt_time) / _benchmark_iterations; opt_time = (get_time() - opt_time) / _benchmark_iterations;
printf ("BenchmarkPsnr_OPT - %8d us opt\n", (int)(opt_time*1e6)); printf("BenchmarkPsnr_OPT - %8.2f us opt\n", opt_time * 1e6);
EXPECT_EQ(0, 0); EXPECT_EQ(0, 0);
@ -274,7 +273,7 @@ TEST_F(libyuvTest, BenchmarkSsim_C) {
_benchmark_width, _benchmark_height); _benchmark_width, _benchmark_height);
c_time = (get_time() - c_time) / _benchmark_iterations; c_time = (get_time() - c_time) / _benchmark_iterations;
printf ("BenchmarkSsim_C - %8d us c\n", (int)(c_time*1e6)); printf("BenchmarkSsim_C - %8.2f us c\n", c_time * 1e6);
MaskCpuFlags(-1); MaskCpuFlags(-1);
@ -297,7 +296,7 @@ TEST_F(libyuvTest, BenchmarkSsim_OPT) {
_benchmark_width, _benchmark_height); _benchmark_width, _benchmark_height);
opt_time = (get_time() - opt_time) / _benchmark_iterations; opt_time = (get_time() - opt_time) / _benchmark_iterations;
printf ("BenchmarkPsnr_OPT - %8d us opt\n", (int)(opt_time*1e6)); printf("BenchmarkPsnr_OPT - %8.2f us opt\n", opt_time * 1e6);
EXPECT_EQ(0, 0); EXPECT_EQ(0, 0);