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

Internals of scale use ptrdiff_t to allow ARGBInterpolate to use stride between 2 images.

Browse Source 
BUG=53
TEST=none
Review URL: https://webrtc-codereview.appspot.com/695004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@306 16f28f9a-4ce2-e073-06de-1de4eb20be90
This commit is contained in:
fbarchard@google.com 2012-07-16 22:58:35 +00:00
parent 4403be58ae
commit 452472dfc4
7 changed files with 137 additions and 109 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: 305 Version: 306
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 305 #define LIBYUV_VERSION 306
#endif // INCLUDE_LIBYUV_VERSION_H_ NOLINT #endif // INCLUDE_LIBYUV_VERSION_H_ NOLINT

14
source/convert.cc
View File

@ -183,15 +183,15 @@ int I422ToI420(const uint8* src_y, int src_stride_y,
// source in scale.cc // source in scale.cc
#if !defined(YUV_DISABLE_ASM) && defined(__ARM_NEON__) #if !defined(YUV_DISABLE_ASM) && defined(__ARM_NEON__)
#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, ptrdiff_t src_stride,
uint8* dst, int dst_width); uint8* dst, int dst_width);
#elif !defined(YUV_DISABLE_ASM) && \ #elif !defined(YUV_DISABLE_ASM) && \
(defined(_M_IX86) || defined(__x86_64__) || defined(__i386__)) (defined(_M_IX86) || defined(__x86_64__) || defined(__i386__))
void ScaleRowDown2Int_SSE2(const uint8* src_ptr, int src_stride, void ScaleRowDown2Int_SSE2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width); uint8* dst_ptr, int dst_width);
#endif #endif
void ScaleRowDown2Int_C(const uint8* src_ptr, int src_stride, void ScaleRowDown2Int_C(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width); uint8* dst_ptr, int dst_width);
int I444ToI420(const uint8* src_y, int src_stride_y, int I444ToI420(const uint8* src_y, int src_stride_y,
@ -212,7 +212,7 @@ int I444ToI420(const uint8* src_y, int src_stride_y,
src_stride_v = -src_stride_v; src_stride_v = -src_stride_v;
} }
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, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) = ScaleRowDown2Int_C; uint8* dst_ptr, int dst_width) = ScaleRowDown2Int_C;
#if defined(HAS_SCALEROWDOWN2_NEON) #if defined(HAS_SCALEROWDOWN2_NEON)
if (TestCpuFlag(kCpuHasNEON) && if (TestCpuFlag(kCpuHasNEON) &&
@ -785,8 +785,10 @@ int UYVYToI420(const uint8* src_uyvy, int src_stride_uyvy,
} }
// Visual C x86 or GCC little endian. // Visual C x86 or GCC little endian.
#if defined(_M_X64) || defined(_M_IX86) || (defined(__BYTE_ORDER) && \ #if defined(__x86_64__) || defined(_M_X64) || \
(__BYTE_ORDER == __ORDER_LITTLE_ENDIAN__ || __BYTE_ORDER == __LITTLE_ENDIAN)) defined(__i386__) || defined(_M_IX86) || \
defined(__arm__) || defined(_M_ARM) || \
(defined(__BYTE_ORDER__) && __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__)
#define LIBYUV_LITTLE_ENDIAN #define LIBYUV_LITTLE_ENDIAN
#endif #endif

6
source/convert_from.cc
View File

@ -402,8 +402,10 @@ static void I42xToUYVYRow_C(const uint8* src_y,
} }
// Visual C x86 or GCC little endian. // Visual C x86 or GCC little endian.
#if defined(_M_X64) || defined(_M_IX86) || (defined(__BYTE_ORDER) && \ #if defined(__x86_64__) || defined(_M_X64) || \
(__BYTE_ORDER == __ORDER_LITTLE_ENDIAN__ || __BYTE_ORDER == __LITTLE_ENDIAN)) defined(__i386__) || defined(_M_IX86) || \
defined(__arm__) || defined(_M_ARM) || \
(defined(__BYTE_ORDER__) && __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__)
#define LIBYUV_LITTLE_ENDIAN #define LIBYUV_LITTLE_ENDIAN
#endif #endif

11
source/planar_functions.cc
View File

@ -1116,11 +1116,12 @@ int ARGBShade(const uint8* src_argb, int src_stride_argb,
(defined(__x86_64__) || defined(__i386__))) (defined(__x86_64__) || defined(__i386__)))
#define HAS_SCALEARGBFILTERROWS_SSSE3 #define HAS_SCALEARGBFILTERROWS_SSSE3
#endif #endif
void ScaleARGBFilterRows_C(uint8* dst_ptr, const uint8* src_ptr, int src_stride, void ScaleARGBFilterRows_C(uint8* dst_ptr,
const uint8* src_ptr, ptrdiff_t src_stride,
int dst_width, int source_y_fraction); int dst_width, int source_y_fraction);
void ScaleARGBFilterRows_SSSE3(uint8* dst_ptr, const uint8* src_ptr, void ScaleARGBFilterRows_SSSE3(uint8* dst_ptr,
int src_stride, int dst_width, const uint8* src_ptr, ptrdiff_t src_stride,
int source_y_fraction); int dst_width, int source_y_fraction);
// Interpolate 2 ARGB images by specified amount (0 to 255). // Interpolate 2 ARGB images by specified amount (0 to 255).
int ARGBInterpolate(const uint8* src_argb0, int src_stride_argb0, int ARGBInterpolate(const uint8* src_argb0, int src_stride_argb0,
@ -1137,7 +1138,7 @@ int ARGBInterpolate(const uint8* src_argb0, int src_stride_argb0,
dst_stride_argb = -dst_stride_argb; dst_stride_argb = -dst_stride_argb;
} }
void (*ScaleARGBFilterRows)(uint8* dst_ptr, const uint8* src_ptr, void (*ScaleARGBFilterRows)(uint8* dst_ptr, const uint8* src_ptr,
int src_stride, int dst_width, ptrdiff_t src_stride, int dst_width,
int source_y_fraction) = ScaleARGBFilterRows_C; int source_y_fraction) = ScaleARGBFilterRows_C;
#if defined(HAS_SCALEARGBFILTERROWS_SSSE3) #if defined(HAS_SCALEARGBFILTERROWS_SSSE3)
if (TestCpuFlag(kCpuHasSSSE3) && if (TestCpuFlag(kCpuHasSSSE3) &&

162
source/scale.cc
View File

@ -52,7 +52,7 @@ void SetUseReferenceImpl(bool use) {
#if !defined(YUV_DISABLE_ASM) && defined(__ARM_NEON__) #if !defined(YUV_DISABLE_ASM) && defined(__ARM_NEON__)
#define HAS_SCALEROWDOWN2_NEON #define HAS_SCALEROWDOWN2_NEON
// Note - not static due to reuse in convert for 444 to 420. // Note - not static due to reuse in convert for 444 to 420.
void ScaleRowDown2_NEON(const uint8* src_ptr, int /* src_stride */, void ScaleRowDown2_NEON(const uint8* src_ptr, ptrdiff_t /* src_stride */,
uint8* dst, int dst_width) { uint8* dst, int dst_width) {
asm volatile ( asm volatile (
"1: \n" "1: \n"
@ -69,7 +69,7 @@ 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, ptrdiff_t src_stride,
uint8* dst, int dst_width) { uint8* dst, int dst_width) {
asm volatile ( asm volatile (
// change the stride to row 2 pointer // change the stride to row 2 pointer
@ -97,7 +97,7 @@ 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, ptrdiff_t /* src_stride */,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile ( asm volatile (
"1: \n" "1: \n"
@ -116,7 +116,7 @@ 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, ptrdiff_t 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"
@ -155,7 +155,8 @@ static void ScaleRowDown4Int_NEON(const uint8* src_ptr, int src_stride,
// Down scale from 4 to 3 pixels. Use the neon multilane read/write // Down scale from 4 to 3 pixels. Use the neon multilane read/write
// to load up the every 4th pixel into a 4 different registers. // to load up the every 4th pixel into a 4 different registers.
// 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,
ptrdiff_t /* src_stride */,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile ( asm volatile (
"1: \n" "1: \n"
@ -172,7 +173,8 @@ 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,
ptrdiff_t 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"
@ -227,7 +229,8 @@ 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,
ptrdiff_t 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"
@ -279,7 +282,8 @@ const vec16 kMult38_Div9 =
65536 / 18, 65536 / 18, 65536 / 18, 65536 / 18 }; 65536 / 18, 65536 / 18, 65536 / 18, 65536 / 18 };
// 32 -> 12 // 32 -> 12
static void ScaleRowDown38_NEON(const uint8* src_ptr, int, static void ScaleRowDown38_NEON(const uint8* src_ptr,
ptrdiff_t /* src_stride */,
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"
@ -301,7 +305,7 @@ static void ScaleRowDown38_NEON(const uint8* src_ptr, int,
// 32x3 -> 12x1 // 32x3 -> 12x1
static void OMITFP ScaleRowDown38_3_Int_NEON(const uint8* src_ptr, static void OMITFP ScaleRowDown38_3_Int_NEON(const uint8* src_ptr,
int src_stride, ptrdiff_t 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"
@ -410,7 +414,8 @@ static void OMITFP ScaleRowDown38_3_Int_NEON(const uint8* src_ptr,
} }
// 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,
ptrdiff_t 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"
@ -505,7 +510,7 @@ static void ScaleRowDown38_2_Int_NEON(const uint8* src_ptr, int src_stride,
// 16x2 -> 16x1 // 16x2 -> 16x1
#define HAS_SCALEFILTERROWS_NEON #define HAS_SCALEFILTERROWS_NEON
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, ptrdiff_t 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"
@ -658,7 +663,7 @@ CONST uvec16 kScaleAb2 =
// Reads 32 pixels, throws half away and writes 16 pixels. // Reads 32 pixels, throws half away and writes 16 pixels.
// Alignment requirement: src_ptr 16 byte aligned, dst_ptr 16 byte aligned. // Alignment requirement: src_ptr 16 byte aligned, dst_ptr 16 byte aligned.
__declspec(naked) __declspec(align(16)) __declspec(naked) __declspec(align(16))
static void ScaleRowDown2_SSE2(const uint8* src_ptr, int src_stride, static void ScaleRowDown2_SSE2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
__asm { __asm {
mov eax, [esp + 4] // src_ptr mov eax, [esp + 4] // src_ptr
@ -687,7 +692,7 @@ static void ScaleRowDown2_SSE2(const uint8* src_ptr, int src_stride,
// Blends 32x2 rectangle to 16x1. // Blends 32x2 rectangle to 16x1.
// Alignment requirement: src_ptr 16 byte aligned, dst_ptr 16 byte aligned. // Alignment requirement: src_ptr 16 byte aligned, dst_ptr 16 byte aligned.
__declspec(naked) __declspec(align(16)) __declspec(naked) __declspec(align(16))
void ScaleRowDown2Int_SSE2(const uint8* src_ptr, int src_stride, void ScaleRowDown2Int_SSE2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
__asm { __asm {
push esi push esi
@ -731,7 +736,8 @@ void ScaleRowDown2Int_SSE2(const uint8* src_ptr, int src_stride,
// Reads 32 pixels, throws half away and writes 16 pixels. // Reads 32 pixels, throws half away and writes 16 pixels.
// Alignment requirement: src_ptr 16 byte aligned, dst_ptr 16 byte aligned. // Alignment requirement: src_ptr 16 byte aligned, dst_ptr 16 byte aligned.
__declspec(naked) __declspec(align(16)) __declspec(naked) __declspec(align(16))
static void ScaleRowDown2_Unaligned_SSE2(const uint8* src_ptr, int src_stride, static void ScaleRowDown2_Unaligned_SSE2(const uint8* src_ptr,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
__asm { __asm {
mov eax, [esp + 4] // src_ptr mov eax, [esp + 4] // src_ptr
@ -761,7 +767,7 @@ static void ScaleRowDown2_Unaligned_SSE2(const uint8* src_ptr, int src_stride,
// Alignment requirement: src_ptr 16 byte aligned, dst_ptr 16 byte aligned. // Alignment requirement: src_ptr 16 byte aligned, dst_ptr 16 byte aligned.
__declspec(naked) __declspec(align(16)) __declspec(naked) __declspec(align(16))
static void ScaleRowDown2Int_Unaligned_SSE2(const uint8* src_ptr, static void ScaleRowDown2Int_Unaligned_SSE2(const uint8* src_ptr,
int src_stride, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
__asm { __asm {
push esi push esi
@ -806,7 +812,7 @@ static void ScaleRowDown2Int_Unaligned_SSE2(const uint8* src_ptr,
// Point samples 32 pixels to 8 pixels. // Point samples 32 pixels to 8 pixels.
// Alignment requirement: src_ptr 16 byte aligned, dst_ptr 8 byte aligned. // Alignment requirement: src_ptr 16 byte aligned, dst_ptr 8 byte aligned.
__declspec(naked) __declspec(align(16)) __declspec(naked) __declspec(align(16))
static void ScaleRowDown4_SSE2(const uint8* src_ptr, int src_stride, static void ScaleRowDown4_SSE2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
__asm { __asm {
mov eax, [esp + 4] // src_ptr mov eax, [esp + 4] // src_ptr
@ -837,7 +843,7 @@ static void ScaleRowDown4_SSE2(const uint8* src_ptr, int src_stride,
// Blends 32x4 rectangle to 8x1. // Blends 32x4 rectangle to 8x1.
// Alignment requirement: src_ptr 16 byte aligned, dst_ptr 8 byte aligned. // Alignment requirement: src_ptr 16 byte aligned, dst_ptr 8 byte aligned.
__declspec(naked) __declspec(align(16)) __declspec(naked) __declspec(align(16))
static void ScaleRowDown4Int_SSE2(const uint8* src_ptr, int src_stride, static void ScaleRowDown4Int_SSE2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
__asm { __asm {
push esi push esi
@ -899,7 +905,7 @@ static void ScaleRowDown4Int_SSE2(const uint8* src_ptr, int src_stride,
// Point samples 32 pixels to 4 pixels. // Point samples 32 pixels to 4 pixels.
// Alignment requirement: src_ptr 16 byte aligned, dst_ptr 4 byte aligned. // Alignment requirement: src_ptr 16 byte aligned, dst_ptr 4 byte aligned.
__declspec(naked) __declspec(align(16)) __declspec(naked) __declspec(align(16))
static void ScaleRowDown8_SSE2(const uint8* src_ptr, int src_stride, static void ScaleRowDown8_SSE2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
__asm { __asm {
mov eax, [esp + 4] // src_ptr mov eax, [esp + 4] // src_ptr
@ -931,7 +937,7 @@ static void ScaleRowDown8_SSE2(const uint8* src_ptr, int src_stride,
// Blends 32x8 rectangle to 4x1. // Blends 32x8 rectangle to 4x1.
// Alignment requirement: src_ptr 16 byte aligned, dst_ptr 4 byte aligned. // Alignment requirement: src_ptr 16 byte aligned, dst_ptr 4 byte aligned.
__declspec(naked) __declspec(align(16)) __declspec(naked) __declspec(align(16))
static void ScaleRowDown8Int_SSE2(const uint8* src_ptr, int src_stride, static void ScaleRowDown8Int_SSE2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
__asm { __asm {
push esi push esi
@ -1009,7 +1015,7 @@ static void ScaleRowDown8Int_SSE2(const uint8* src_ptr, int src_stride,
// Note that movdqa+palign may be better than movdqu. // Note that movdqa+palign may be better than movdqu.
// Alignment requirement: src_ptr 16 byte aligned, dst_ptr 8 byte aligned. // Alignment requirement: src_ptr 16 byte aligned, dst_ptr 8 byte aligned.
__declspec(naked) __declspec(align(16)) __declspec(naked) __declspec(align(16))
static void ScaleRowDown34_SSSE3(const uint8* src_ptr, int src_stride, static void ScaleRowDown34_SSSE3(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
__asm { __asm {
mov eax, [esp + 4] // src_ptr mov eax, [esp + 4] // src_ptr
@ -1058,7 +1064,8 @@ static void ScaleRowDown34_SSSE3(const uint8* src_ptr, int src_stride,
// Note that movdqa+palign may be better than movdqu. // Note that movdqa+palign may be better than movdqu.
// Alignment requirement: src_ptr 16 byte aligned, dst_ptr 8 byte aligned. // Alignment requirement: src_ptr 16 byte aligned, dst_ptr 8 byte aligned.
__declspec(naked) __declspec(align(16)) __declspec(naked) __declspec(align(16))
static void ScaleRowDown34_1_Int_SSSE3(const uint8* src_ptr, int src_stride, static void ScaleRowDown34_1_Int_SSSE3(const uint8* src_ptr,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
__asm { __asm {
push esi push esi
@ -1116,7 +1123,8 @@ static void ScaleRowDown34_1_Int_SSSE3(const uint8* src_ptr, int src_stride,
// Note that movdqa+palign may be better than movdqu. // Note that movdqa+palign may be better than movdqu.
// Alignment requirement: src_ptr 16 byte aligned, dst_ptr 8 byte aligned. // Alignment requirement: src_ptr 16 byte aligned, dst_ptr 8 byte aligned.
__declspec(naked) __declspec(align(16)) __declspec(naked) __declspec(align(16))
static void ScaleRowDown34_0_Int_SSSE3(const uint8* src_ptr, int src_stride, static void ScaleRowDown34_0_Int_SSSE3(const uint8* src_ptr,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
__asm { __asm {
push esi push esi
@ -1179,7 +1187,7 @@ static void ScaleRowDown34_0_Int_SSSE3(const uint8* src_ptr, int src_stride,
// Scale 32 pixels to 12 // Scale 32 pixels to 12
__declspec(naked) __declspec(align(16)) __declspec(naked) __declspec(align(16))
static void ScaleRowDown38_SSSE3(const uint8* src_ptr, int src_stride, static void ScaleRowDown38_SSSE3(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
__asm { __asm {
mov eax, [esp + 4] // src_ptr mov eax, [esp + 4] // src_ptr
@ -1211,7 +1219,8 @@ static void ScaleRowDown38_SSSE3(const uint8* src_ptr, int src_stride,
// Scale 16x3 pixels to 6x1 with interpolation // Scale 16x3 pixels to 6x1 with interpolation
__declspec(naked) __declspec(align(16)) __declspec(naked) __declspec(align(16))
static void ScaleRowDown38_3_Int_SSSE3(const uint8* src_ptr, int src_stride, static void ScaleRowDown38_3_Int_SSSE3(const uint8* src_ptr,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
__asm { __asm {
push esi push esi
@ -1276,7 +1285,8 @@ static void ScaleRowDown38_3_Int_SSSE3(const uint8* src_ptr, int src_stride,
// Scale 16x2 pixels to 6x1 with interpolation // Scale 16x2 pixels to 6x1 with interpolation
__declspec(naked) __declspec(align(16)) __declspec(naked) __declspec(align(16))
static void ScaleRowDown38_2_Int_SSSE3(const uint8* src_ptr, int src_stride, static void ScaleRowDown38_2_Int_SSSE3(const uint8* src_ptr,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
__asm { __asm {
push esi push esi
@ -1322,7 +1332,7 @@ static void ScaleRowDown38_2_Int_SSSE3(const uint8* src_ptr, int src_stride,
// Reads 16xN bytes and produces 16 shorts at a time. // Reads 16xN bytes and produces 16 shorts at a time.
__declspec(naked) __declspec(align(16)) __declspec(naked) __declspec(align(16))
static void ScaleAddRows_SSE2(const uint8* src_ptr, int src_stride, static void ScaleAddRows_SSE2(const uint8* src_ptr, ptrdiff_t src_stride,
uint16* dst_ptr, int src_width, uint16* dst_ptr, int src_width,
int src_height) { int src_height) {
__asm { __asm {
@ -1388,7 +1398,7 @@ static void ScaleAddRows_SSE2(const uint8* src_ptr, int src_stride,
#define HAS_SCALEFILTERROWS_SSE2_DISABLED #define HAS_SCALEFILTERROWS_SSE2_DISABLED
__declspec(naked) __declspec(align(16)) __declspec(naked) __declspec(align(16))
static void ScaleFilterRows_SSE2(uint8* dst_ptr, const uint8* src_ptr, static void ScaleFilterRows_SSE2(uint8* dst_ptr, const uint8* src_ptr,
int src_stride, int dst_width, ptrdiff_t src_stride, int dst_width,
int source_y_fraction) { int source_y_fraction) {
__asm { __asm {
push esi push esi
@ -1479,7 +1489,7 @@ static void ScaleFilterRows_SSE2(uint8* dst_ptr, const uint8* src_ptr,
#define HAS_SCALEFILTERROWS_SSSE3 #define HAS_SCALEFILTERROWS_SSSE3
__declspec(naked) __declspec(align(16)) __declspec(naked) __declspec(align(16))
static void ScaleFilterRows_SSSE3(uint8* dst_ptr, const uint8* src_ptr, static void ScaleFilterRows_SSSE3(uint8* dst_ptr, const uint8* src_ptr,
int src_stride, int dst_width, ptrdiff_t src_stride, int dst_width,
int source_y_fraction) { int source_y_fraction) {
__asm { __asm {
push esi push esi
@ -1570,7 +1580,7 @@ static void ScaleFilterRows_SSSE3(uint8* dst_ptr, const uint8* src_ptr,
// Generated using gcc disassembly on Visual C object file: // Generated using gcc disassembly on Visual C object file:
// objdump -D yuvscaler.obj >yuvscaler.txt // objdump -D yuvscaler.obj >yuvscaler.txt
#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, ptrdiff_t 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"
@ -1598,7 +1608,7 @@ static void ScaleRowDown2_SSE2(const uint8* src_ptr, int src_stride,
); );
} }
void ScaleRowDown2Int_SSE2(const uint8* src_ptr, int src_stride, void ScaleRowDown2Int_SSE2(const uint8* src_ptr, ptrdiff_t 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"
@ -1635,7 +1645,8 @@ void ScaleRowDown2Int_SSE2(const uint8* src_ptr, int src_stride,
#endif #endif
); );
} }
static void ScaleRowDown2_Unaligned_SSE2(const uint8* src_ptr, int src_stride, static void ScaleRowDown2_Unaligned_SSE2(const uint8* src_ptr,
ptrdiff_t 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"
@ -1664,7 +1675,7 @@ static void ScaleRowDown2_Unaligned_SSE2(const uint8* src_ptr, int src_stride,
} }
static void ScaleRowDown2Int_Unaligned_SSE2(const uint8* src_ptr, static void ScaleRowDown2Int_Unaligned_SSE2(const uint8* src_ptr,
int src_stride, ptrdiff_t 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"
@ -1703,7 +1714,7 @@ static void ScaleRowDown2Int_Unaligned_SSE2(const uint8* src_ptr,
} }
#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, ptrdiff_t 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"
@ -1732,7 +1743,7 @@ 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, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
intptr_t stridex3 = 0; intptr_t stridex3 = 0;
asm volatile ( asm volatile (
@ -1787,7 +1798,7 @@ 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, ptrdiff_t 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"
@ -1817,7 +1828,7 @@ static void ScaleRowDown8_SSE2(const uint8* src_ptr, int src_stride,
); );
} }
static void ScaleRowDown8Int_SSE2(const uint8* src_ptr, int src_stride, static void ScaleRowDown8Int_SSE2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
intptr_t stridex3 = 0; intptr_t stridex3 = 0;
intptr_t row4 = 0; intptr_t row4 = 0;
@ -1884,7 +1895,7 @@ 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, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile ( asm volatile (
"movdqa %0,%%xmm3 \n" "movdqa %0,%%xmm3 \n"
@ -1923,7 +1934,8 @@ 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,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile ( asm volatile (
"movdqa %0,%%xmm2 \n" // kShuf01 "movdqa %0,%%xmm2 \n" // kShuf01
@ -1989,7 +2001,8 @@ 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,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile ( asm volatile (
"movdqa %0,%%xmm2 \n" // kShuf01 "movdqa %0,%%xmm2 \n" // kShuf01
@ -2060,7 +2073,7 @@ 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, ptrdiff_t 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"
@ -2091,7 +2104,8 @@ static void ScaleRowDown38_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,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile ( asm volatile (
"movdqa %0,%%xmm2 \n" "movdqa %0,%%xmm2 \n"
@ -2136,7 +2150,8 @@ static void ScaleRowDown38_2_Int_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,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile ( asm volatile (
"movdqa %0,%%xmm2 \n" "movdqa %0,%%xmm2 \n"
@ -2201,7 +2216,7 @@ static void ScaleRowDown38_3_Int_SSSE3(const uint8* src_ptr, int src_stride,
} }
#define HAS_SCALEADDROWS_SSE2 #define HAS_SCALEADDROWS_SSE2
static void ScaleAddRows_SSE2(const uint8* src_ptr, int src_stride, static void ScaleAddRows_SSE2(const uint8* src_ptr, ptrdiff_t 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;
@ -2254,7 +2269,7 @@ static void ScaleAddRows_SSE2(const uint8* src_ptr, int src_stride,
// Bilinear row filtering combines 16x2 -> 16x1. SSE2 version // Bilinear row filtering combines 16x2 -> 16x1. SSE2 version
#define HAS_SCALEFILTERROWS_SSE2_DISABLED #define HAS_SCALEFILTERROWS_SSE2_DISABLED
static void ScaleFilterRows_SSE2(uint8* dst_ptr, static void ScaleFilterRows_SSE2(uint8* dst_ptr,
const uint8* src_ptr, int src_stride, const uint8* src_ptr, ptrdiff_t src_stride,
int dst_width, int source_y_fraction) { int dst_width, int source_y_fraction) {
asm volatile ( asm volatile (
"sub %1,%0 \n" "sub %1,%0 \n"
@ -2327,7 +2342,7 @@ static void ScaleFilterRows_SSE2(uint8* dst_ptr,
// Bilinear row filtering combines 16x2 -> 16x1. SSSE3 version // Bilinear row filtering combines 16x2 -> 16x1. SSSE3 version
#define HAS_SCALEFILTERROWS_SSSE3 #define HAS_SCALEFILTERROWS_SSSE3
static void ScaleFilterRows_SSSE3(uint8* dst_ptr, static void ScaleFilterRows_SSSE3(uint8* dst_ptr,
const uint8* src_ptr, int src_stride, const uint8* src_ptr, ptrdiff_t src_stride,
int dst_width, int source_y_fraction) { int dst_width, int source_y_fraction) {
asm volatile ( asm volatile (
"sub %1,%0 \n" "sub %1,%0 \n"
@ -2396,7 +2411,7 @@ static void ScaleFilterRows_SSSE3(uint8* dst_ptr,
#endif // defined(__x86_64__) || defined(__i386__) #endif // defined(__x86_64__) || defined(__i386__)
// CPU agnostic row functions // CPU agnostic row functions
static void ScaleRowDown2_C(const uint8* src_ptr, int, static void ScaleRowDown2_C(const uint8* src_ptr, ptrdiff_t /* src_stride */,
uint8* dst, int dst_width) { uint8* dst, int dst_width) {
uint8* dend = dst + dst_width - 1; uint8* dend = dst + dst_width - 1;
do { do {
@ -2410,7 +2425,7 @@ static void ScaleRowDown2_C(const uint8* src_ptr, int,
} }
} }
void ScaleRowDown2Int_C(const uint8* src_ptr, int src_stride, void ScaleRowDown2Int_C(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width) { uint8* dst, int dst_width) {
const uint8* s = src_ptr; const uint8* s = src_ptr;
const uint8* t = src_ptr + src_stride; const uint8* t = src_ptr + src_stride;
@ -2427,7 +2442,7 @@ void ScaleRowDown2Int_C(const uint8* src_ptr, int src_stride,
} }
} }
static void ScaleRowDown4_C(const uint8* src_ptr, int, static void ScaleRowDown4_C(const uint8* src_ptr, ptrdiff_t /* src_stride */,
uint8* dst, int dst_width) { uint8* dst, int dst_width) {
uint8* dend = dst + dst_width - 1; uint8* dend = dst + dst_width - 1;
do { do {
@ -2441,7 +2456,7 @@ static void ScaleRowDown4_C(const uint8* src_ptr, int,
} }
} }
static void ScaleRowDown4Int_C(const uint8* src_ptr, int src_stride, static void ScaleRowDown4Int_C(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width) { uint8* dst, int dst_width) {
intptr_t stride = src_stride; intptr_t stride = src_stride;
uint8* dend = dst + dst_width - 1; uint8* dend = dst + dst_width - 1;
@ -2482,7 +2497,7 @@ static void ScaleRowDown4Int_C(const uint8* src_ptr, int src_stride,
static const int kMaxOutputWidth = 640; static const int kMaxOutputWidth = 640;
static const int kMaxRow12 = kMaxOutputWidth * 2; static const int kMaxRow12 = kMaxOutputWidth * 2;
static void ScaleRowDown8_C(const uint8* src_ptr, int, static void ScaleRowDown8_C(const uint8* src_ptr, ptrdiff_t /* src_stride */,
uint8* dst, int dst_width) { uint8* dst, int dst_width) {
uint8* dend = dst + dst_width - 1; uint8* dend = dst + dst_width - 1;
do { do {
@ -2498,7 +2513,7 @@ static void ScaleRowDown8_C(const uint8* src_ptr, int,
// Note calling code checks width is less than max and if not // Note calling code checks width is less than max and if not
// uses ScaleRowDown8_C instead. // uses ScaleRowDown8_C instead.
static void ScaleRowDown8Int_C(const uint8* src_ptr, int src_stride, static void ScaleRowDown8Int_C(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width) { uint8* dst, int dst_width) {
SIMD_ALIGNED(uint8 src_row[kMaxRow12 * 2]); SIMD_ALIGNED(uint8 src_row[kMaxRow12 * 2]);
assert(dst_width <= kMaxOutputWidth); assert(dst_width <= kMaxOutputWidth);
@ -2509,7 +2524,7 @@ static void ScaleRowDown8Int_C(const uint8* src_ptr, int src_stride,
ScaleRowDown2Int_C(src_row, kMaxOutputWidth, dst, dst_width); ScaleRowDown2Int_C(src_row, kMaxOutputWidth, dst, dst_width);
} }
static void ScaleRowDown34_C(const uint8* src_ptr, int, static void ScaleRowDown34_C(const uint8* src_ptr, ptrdiff_t /* src_stride */,
uint8* dst, int dst_width) { uint8* dst, int dst_width) {
assert((dst_width % 3 == 0) && (dst_width > 0)); assert((dst_width % 3 == 0) && (dst_width > 0));
uint8* dend = dst + dst_width; uint8* dend = dst + dst_width;
@ -2523,7 +2538,7 @@ static void ScaleRowDown34_C(const uint8* src_ptr, int,
} }
// Filter rows 0 and 1 together, 3 : 1 // Filter rows 0 and 1 together, 3 : 1
static void ScaleRowDown34_0_Int_C(const uint8* src_ptr, int src_stride, static void ScaleRowDown34_0_Int_C(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* d, int dst_width) { uint8* d, int dst_width) {
assert((dst_width % 3 == 0) && (dst_width > 0)); assert((dst_width % 3 == 0) && (dst_width > 0));
const uint8* s = src_ptr; const uint8* s = src_ptr;
@ -2546,7 +2561,7 @@ static void ScaleRowDown34_0_Int_C(const uint8* src_ptr, int src_stride,
} }
// Filter rows 1 and 2 together, 1 : 1 // Filter rows 1 and 2 together, 1 : 1
static void ScaleRowDown34_1_Int_C(const uint8* src_ptr, int src_stride, static void ScaleRowDown34_1_Int_C(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* d, int dst_width) { uint8* d, int dst_width) {
assert((dst_width % 3 == 0) && (dst_width > 0)); assert((dst_width % 3 == 0) && (dst_width > 0));
const uint8* s = src_ptr; const uint8* s = src_ptr;
@ -2615,7 +2630,8 @@ static void ScaleFilterCols34_C(uint8* dst_ptr, const uint8* src_ptr,
#define HAS_SCALEROWDOWN34_SSE2_DISABLED #define HAS_SCALEROWDOWN34_SSE2_DISABLED
// Filter rows 0 and 1 together, 3 : 1 // Filter rows 0 and 1 together, 3 : 1
static void ScaleRowDown34_0_Int_SSE2(const uint8* src_ptr, int src_stride, static void ScaleRowDown34_0_Int_SSE2(const uint8* src_ptr,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
assert((dst_width % 3 == 0) && (dst_width > 0)); assert((dst_width % 3 == 0) && (dst_width > 0));
SIMD_ALIGNED(uint8 row[kMaxInputWidth]); SIMD_ALIGNED(uint8 row[kMaxInputWidth]);
@ -2624,7 +2640,8 @@ static void ScaleRowDown34_0_Int_SSE2(const uint8* src_ptr, int src_stride,
} }
// Filter rows 1 and 2 together, 1 : 1 // Filter rows 1 and 2 together, 1 : 1
static void ScaleRowDown34_1_Int_SSE2(const uint8* src_ptr, int src_stride, static void ScaleRowDown34_1_Int_SSE2(const uint8* src_ptr,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
assert((dst_width % 3 == 0) && (dst_width > 0)); assert((dst_width % 3 == 0) && (dst_width > 0));
SIMD_ALIGNED(uint8 row[kMaxInputWidth]); SIMD_ALIGNED(uint8 row[kMaxInputWidth]);
@ -2633,7 +2650,7 @@ static void ScaleRowDown34_1_Int_SSE2(const uint8* src_ptr, int src_stride,
} }
#endif #endif
static void ScaleRowDown38_C(const uint8* src_ptr, int, static void ScaleRowDown38_C(const uint8* src_ptr, ptrdiff_t /* src_stride */,
uint8* dst, int dst_width) { uint8* dst, int dst_width) {
assert(dst_width % 3 == 0); assert(dst_width % 3 == 0);
for (int x = 0; x < dst_width; x += 3) { for (int x = 0; x < dst_width; x += 3) {
@ -2646,7 +2663,8 @@ static void ScaleRowDown38_C(const uint8* src_ptr, int,
} }
// 8x3 -> 3x1 // 8x3 -> 3x1
static void ScaleRowDown38_3_Int_C(const uint8* src_ptr, int src_stride, static void ScaleRowDown38_3_Int_C(const uint8* src_ptr,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
assert((dst_width % 3 == 0) && (dst_width > 0)); assert((dst_width % 3 == 0) && (dst_width > 0));
intptr_t stride = src_stride; intptr_t stride = src_stride;
@ -2671,7 +2689,7 @@ static void ScaleRowDown38_3_Int_C(const uint8* src_ptr, int src_stride,
} }
// 8x2 -> 3x1 // 8x2 -> 3x1
static void ScaleRowDown38_2_Int_C(const uint8* src_ptr, int src_stride, static void ScaleRowDown38_2_Int_C(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
assert((dst_width % 3 == 0) && (dst_width > 0)); assert((dst_width % 3 == 0) && (dst_width > 0));
intptr_t stride = src_stride; intptr_t stride = src_stride;
@ -2692,7 +2710,7 @@ static void ScaleRowDown38_2_Int_C(const uint8* src_ptr, int src_stride,
// C version 8x2 -> 8x1 // C version 8x2 -> 8x1
static void ScaleFilterRows_C(uint8* dst_ptr, static void ScaleFilterRows_C(uint8* dst_ptr,
const uint8* src_ptr, int src_stride, const uint8* src_ptr, ptrdiff_t src_stride,
int dst_width, int source_y_fraction) { int dst_width, int source_y_fraction) {
assert(dst_width > 0); assert(dst_width > 0);
int y1_fraction = source_y_fraction; int y1_fraction = source_y_fraction;
@ -2715,7 +2733,7 @@ static void ScaleFilterRows_C(uint8* dst_ptr,
dst_ptr[0] = dst_ptr[-1]; dst_ptr[0] = dst_ptr[-1];
} }
void ScaleAddRows_C(const uint8* src_ptr, int src_stride, void ScaleAddRows_C(const uint8* src_ptr, ptrdiff_t src_stride,
uint16* dst_ptr, int src_width, int src_height) { uint16* dst_ptr, int src_width, int src_height) {
assert(src_width > 0); assert(src_width > 0);
assert(src_height > 0); assert(src_height > 0);
@ -2744,7 +2762,7 @@ static void ScalePlaneDown2(int src_width, int src_height,
FilterMode filtering) { FilterMode filtering) {
assert(IS_ALIGNED(src_width, 2)); assert(IS_ALIGNED(src_width, 2));
assert(IS_ALIGNED(src_height, 2)); assert(IS_ALIGNED(src_height, 2));
void (*ScaleRowDown2)(const uint8* src_ptr, int src_stride, void (*ScaleRowDown2)(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) = uint8* dst_ptr, int dst_width) =
filtering ? ScaleRowDown2Int_C : ScaleRowDown2_C; filtering ? ScaleRowDown2Int_C : ScaleRowDown2_C;
#if defined(HAS_SCALEROWDOWN2_NEON) #if defined(HAS_SCALEROWDOWN2_NEON)
@ -2784,7 +2802,7 @@ static void ScalePlaneDown4(int src_width, int src_height,
FilterMode filtering) { FilterMode filtering) {
assert(IS_ALIGNED(src_width, 4)); assert(IS_ALIGNED(src_width, 4));
assert(IS_ALIGNED(src_height, 4)); assert(IS_ALIGNED(src_height, 4));
void (*ScaleRowDown4)(const uint8* src_ptr, int src_stride, void (*ScaleRowDown4)(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) = uint8* dst_ptr, int dst_width) =
filtering ? ScaleRowDown4Int_C : ScaleRowDown4_C; filtering ? ScaleRowDown4Int_C : ScaleRowDown4_C;
#if defined(HAS_SCALEROWDOWN4_NEON) #if defined(HAS_SCALEROWDOWN4_NEON)
@ -2821,7 +2839,7 @@ static void ScalePlaneDown8(int src_width, int src_height,
FilterMode filtering) { FilterMode filtering) {
assert(IS_ALIGNED(src_width, 8)); assert(IS_ALIGNED(src_width, 8));
assert(IS_ALIGNED(src_height, 8)); assert(IS_ALIGNED(src_height, 8));
void (*ScaleRowDown8)(const uint8* src_ptr, int src_stride, void (*ScaleRowDown8)(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) = uint8* dst_ptr, int dst_width) =
filtering && (dst_width <= kMaxOutputWidth) ? filtering && (dst_width <= kMaxOutputWidth) ?
ScaleRowDown8Int_C : ScaleRowDown8_C; ScaleRowDown8Int_C : ScaleRowDown8_C;
@ -2852,9 +2870,9 @@ static void ScalePlaneDown34(int src_width, int src_height,
const uint8* src_ptr, uint8* dst_ptr, const uint8* src_ptr, uint8* dst_ptr,
FilterMode filtering) { FilterMode filtering) {
assert(dst_width % 3 == 0); assert(dst_width % 3 == 0);
void (*ScaleRowDown34_0)(const uint8* src_ptr, int src_stride, void (*ScaleRowDown34_0)(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width); uint8* dst_ptr, int dst_width);
void (*ScaleRowDown34_1)(const uint8* src_ptr, int src_stride, void (*ScaleRowDown34_1)(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width); uint8* dst_ptr, int dst_width);
if (!filtering) { if (!filtering) {
ScaleRowDown34_0 = ScaleRowDown34_C; ScaleRowDown34_0 = ScaleRowDown34_C;
@ -2941,9 +2959,9 @@ static void ScalePlaneDown38(int src_width, int src_height,
const uint8* src_ptr, uint8* dst_ptr, const uint8* src_ptr, uint8* dst_ptr,
FilterMode filtering) { FilterMode filtering) {
assert(dst_width % 3 == 0); assert(dst_width % 3 == 0);
void (*ScaleRowDown38_3)(const uint8* src_ptr, int src_stride, void (*ScaleRowDown38_3)(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width); uint8* dst_ptr, int dst_width);
void (*ScaleRowDown38_2)(const uint8* src_ptr, int src_stride, void (*ScaleRowDown38_2)(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width); uint8* dst_ptr, int dst_width);
if (!filtering) { if (!filtering) {
ScaleRowDown38_3 = ScaleRowDown38_C; ScaleRowDown38_3 = ScaleRowDown38_C;
@ -2999,7 +3017,7 @@ static void ScalePlaneDown38(int src_width, int src_height,
} }
static __inline uint32 SumBox(int iboxwidth, int iboxheight, static __inline uint32 SumBox(int iboxwidth, int iboxheight,
int src_stride, const uint8* src_ptr) { ptrdiff_t src_stride, const uint8* src_ptr) {
assert(iboxwidth > 0); assert(iboxwidth > 0);
assert(iboxheight > 0); assert(iboxheight > 0);
uint32 sum = 0u; uint32 sum = 0u;
@ -3013,7 +3031,7 @@ static __inline uint32 SumBox(int iboxwidth, int iboxheight,
} }
static void ScalePlaneBoxRow_C(int dst_width, int boxheight, static void ScalePlaneBoxRow_C(int dst_width, int boxheight,
int x, int dx, int src_stride, int x, int dx, ptrdiff_t src_stride,
const uint8* src_ptr, uint8* dst_ptr) { const uint8* src_ptr, uint8* dst_ptr) {
for (int i = 0; i < dst_width; ++i) { for (int i = 0; i < dst_width; ++i) {
int ix = x >> 16; int ix = x >> 16;
@ -3096,7 +3114,7 @@ static void ScalePlaneBox(int src_width, int src_height,
} }
} else { } else {
SIMD_ALIGNED(uint16 row[kMaxInputWidth]); SIMD_ALIGNED(uint16 row[kMaxInputWidth]);
void (*ScaleAddRows)(const uint8* src_ptr, int src_stride, void (*ScaleAddRows)(const uint8* src_ptr, ptrdiff_t src_stride,
uint16* dst_ptr, int src_width, int src_height)= uint16* dst_ptr, int src_width, int src_height)=
ScaleAddRows_C; ScaleAddRows_C;
void (*ScaleAddCols)(int dst_width, int boxheight, int x, int dx, void (*ScaleAddCols)(int dst_width, int boxheight, int x, int dx,
@ -3186,7 +3204,7 @@ void ScalePlaneBilinear(int src_width, int src_height,
} else { } else {
SIMD_ALIGNED(uint8 row[kMaxInputWidth + 16]); SIMD_ALIGNED(uint8 row[kMaxInputWidth + 16]);
void (*ScaleFilterRows)(uint8* dst_ptr, const uint8* src_ptr, void (*ScaleFilterRows)(uint8* dst_ptr, const uint8* src_ptr,
int src_stride, ptrdiff_t src_stride,
int dst_width, int source_y_fraction) = int dst_width, int source_y_fraction) =
ScaleFilterRows_C; ScaleFilterRows_C;
#if defined(HAS_SCALEFILTERROWS_NEON) #if defined(HAS_SCALEFILTERROWS_NEON)

49
source/scale_argb.cc
View File

@ -37,7 +37,7 @@ extern "C" {
// Reads 8 pixels, throws half away and writes 4 even pixels (0, 2, 4, 6) // Reads 8 pixels, throws half away and writes 4 even pixels (0, 2, 4, 6)
// Alignment requirement: src_ptr 16 byte aligned, dst_ptr 16 byte aligned. // Alignment requirement: src_ptr 16 byte aligned, dst_ptr 16 byte aligned.
__declspec(naked) __declspec(align(16)) __declspec(naked) __declspec(align(16))
static void ScaleARGBRowDown2_SSE2(const uint8* src_ptr, int src_stride, static void ScaleARGBRowDown2_SSE2(const uint8* src_ptr, ptrdiff_t /* src_stride */,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
__asm { __asm {
mov eax, [esp + 4] // src_ptr mov eax, [esp + 4] // src_ptr
@ -63,7 +63,7 @@ static void ScaleARGBRowDown2_SSE2(const uint8* src_ptr, int src_stride,
// Blends 8x2 rectangle to 4x1. // Blends 8x2 rectangle to 4x1.
// Alignment requirement: src_ptr 16 byte aligned, dst_ptr 16 byte aligned. // Alignment requirement: src_ptr 16 byte aligned, dst_ptr 16 byte aligned.
__declspec(naked) __declspec(align(16)) __declspec(naked) __declspec(align(16))
static void ScaleARGBRowDown2Int_SSE2(const uint8* src_ptr, int src_stride, static void ScaleARGBRowDown2Int_SSE2(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
__asm { __asm {
push esi push esi
@ -99,7 +99,7 @@ static void ScaleARGBRowDown2Int_SSE2(const uint8* src_ptr, int src_stride,
// Reads 4 pixels at a time. // Reads 4 pixels at a time.
// Alignment requirement: dst_ptr 16 byte aligned. // Alignment requirement: dst_ptr 16 byte aligned.
__declspec(naked) __declspec(align(16)) __declspec(naked) __declspec(align(16))
void ScaleARGBRowDownEven_SSE2(const uint8* src_ptr, int src_stride, void ScaleARGBRowDownEven_SSE2(const uint8* src_ptr, ptrdiff_t src_stride,
int src_stepx, int src_stepx,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
__asm { __asm {
@ -137,7 +137,7 @@ void ScaleARGBRowDownEven_SSE2(const uint8* src_ptr, int src_stride,
// Blends four 2x2 to 4x1. // Blends four 2x2 to 4x1.
// Alignment requirement: dst_ptr 16 byte aligned. // Alignment requirement: dst_ptr 16 byte aligned.
__declspec(naked) __declspec(align(16)) __declspec(naked) __declspec(align(16))
static void ScaleARGBRowDownEvenInt_SSE2(const uint8* src_ptr, int src_stride, static void ScaleARGBRowDownEvenInt_SSE2(const uint8* src_ptr, ptrdiff_t src_stride,
int src_stepx, int src_stepx,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
__asm { __asm {
@ -188,7 +188,7 @@ static void ScaleARGBRowDownEvenInt_SSE2(const uint8* src_ptr, int src_stride,
#define HAS_SCALEARGBFILTERROWS_SSE2_DISABLED #define HAS_SCALEARGBFILTERROWS_SSE2_DISABLED
__declspec(naked) __declspec(align(16)) __declspec(naked) __declspec(align(16))
void ScaleARGBFilterRows_SSE2(uint8* dst_ptr, const uint8* src_ptr, void ScaleARGBFilterRows_SSE2(uint8* dst_ptr, const uint8* src_ptr,
int src_stride, int dst_width, ptrdiff_t src_stride, int dst_width,
int source_y_fraction) { int source_y_fraction) {
__asm { __asm {
push esi push esi
@ -276,7 +276,7 @@ void ScaleARGBFilterRows_SSE2(uint8* dst_ptr, const uint8* src_ptr,
#define HAS_SCALEARGBFILTERROWS_SSSE3 #define HAS_SCALEARGBFILTERROWS_SSSE3
__declspec(naked) __declspec(align(16)) __declspec(naked) __declspec(align(16))
void ScaleARGBFilterRows_SSSE3(uint8* dst_ptr, const uint8* src_ptr, void ScaleARGBFilterRows_SSSE3(uint8* dst_ptr, const uint8* src_ptr,
int src_stride, int dst_width, ptrdiff_t src_stride, int dst_width,
int source_y_fraction) { int source_y_fraction) {
__asm { __asm {
push esi push esi
@ -360,7 +360,8 @@ void ScaleARGBFilterRows_SSSE3(uint8* dst_ptr, const uint8* src_ptr,
// Generated using gcc disassembly on Visual C object file: // Generated using gcc disassembly on Visual C object file:
// objdump -D yuvscaler.obj >yuvscaler.txt // objdump -D yuvscaler.obj >yuvscaler.txt
#define HAS_SCALEARGBROWDOWN2_SSE2 #define HAS_SCALEARGBROWDOWN2_SSE2
static void ScaleARGBRowDown2_SSE2(const uint8* src_ptr, int , static void ScaleARGBRowDown2_SSE2(const uint8* src_ptr,
ptrdiff_t /* src_stride */,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile ( asm volatile (
".p2align 4 \n" ".p2align 4 \n"
@ -384,7 +385,8 @@ static void ScaleARGBRowDown2_SSE2(const uint8* src_ptr, int ,
); );
} }
static void ScaleARGBRowDown2Int_SSE2(const uint8* src_ptr, int src_stride, static void ScaleARGBRowDown2Int_SSE2(const uint8* src_ptr,
ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
asm volatile ( asm volatile (
".p2align 4 \n" ".p2align 4 \n"
@ -418,7 +420,7 @@ static void ScaleARGBRowDown2Int_SSE2(const uint8* src_ptr, int src_stride,
#define HAS_SCALEARGBROWDOWNEVEN_SSE2 #define HAS_SCALEARGBROWDOWNEVEN_SSE2
// Reads 4 pixels at a time. // Reads 4 pixels at a time.
// Alignment requirement: dst_ptr 16 byte aligned. // Alignment requirement: dst_ptr 16 byte aligned.
void ScaleARGBRowDownEven_SSE2(const uint8* src_ptr, int src_stride, void ScaleARGBRowDownEven_SSE2(const uint8* src_ptr, ptrdiff_t src_stride,
int src_stepx, int src_stepx,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
intptr_t src_stepx_x4 = static_cast<intptr_t>(src_stepx); intptr_t src_stepx_x4 = static_cast<intptr_t>(src_stepx);
@ -455,8 +457,8 @@ void ScaleARGBRowDownEven_SSE2(const uint8* src_ptr, int src_stride,
// Blends four 2x2 to 4x1. // Blends four 2x2 to 4x1.
// Alignment requirement: dst_ptr 16 byte aligned. // Alignment requirement: dst_ptr 16 byte aligned.
static void ScaleARGBRowDownEvenInt_SSE2(const uint8* src_ptr, int src_stride, static void ScaleARGBRowDownEvenInt_SSE2(const uint8* src_ptr,
int src_stepx, ptrdiff_t src_stride, int src_stepx,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
intptr_t src_stepx_x4 = static_cast<intptr_t>(src_stepx); intptr_t src_stepx_x4 = static_cast<intptr_t>(src_stepx);
intptr_t src_stepx_x12 = 0; intptr_t src_stepx_x12 = 0;
@ -505,7 +507,7 @@ static void ScaleARGBRowDownEvenInt_SSE2(const uint8* src_ptr, int src_stride,
// Bilinear row filtering combines 4x2 -> 4x1. SSE2 version // Bilinear row filtering combines 4x2 -> 4x1. SSE2 version
#define HAS_SCALEARGBFILTERROWS_SSE2_DISABLED #define HAS_SCALEARGBFILTERROWS_SSE2_DISABLED
void ScaleARGBFilterRows_SSE2(uint8* dst_ptr, const uint8* src_ptr, void ScaleARGBFilterRows_SSE2(uint8* dst_ptr, const uint8* src_ptr,
int src_stride, int dst_width, ptrdiff_t src_stride, int dst_width,
int source_y_fraction) { int source_y_fraction) {
asm volatile ( asm volatile (
"sub %1,%0 \n" "sub %1,%0 \n"
@ -577,7 +579,7 @@ void ScaleARGBFilterRows_SSE2(uint8* dst_ptr, const uint8* src_ptr,
// Bilinear row filtering combines 4x2 -> 4x1. SSSE3 version // Bilinear row filtering combines 4x2 -> 4x1. SSSE3 version
#define HAS_SCALEARGBFILTERROWS_SSSE3 #define HAS_SCALEARGBFILTERROWS_SSSE3
void ScaleARGBFilterRows_SSSE3(uint8* dst_ptr, const uint8* src_ptr, void ScaleARGBFilterRows_SSSE3(uint8* dst_ptr, const uint8* src_ptr,
int src_stride, int dst_width, ptrdiff_t src_stride, int dst_width,
int source_y_fraction) { int source_y_fraction) {
asm volatile ( asm volatile (
"sub %1,%0 \n" "sub %1,%0 \n"
@ -643,7 +645,8 @@ void ScaleARGBFilterRows_SSSE3(uint8* dst_ptr, const uint8* src_ptr,
} }
#endif // defined(__x86_64__) || defined(__i386__) #endif // defined(__x86_64__) || defined(__i386__)
static void ScaleARGBRowDown2_C(const uint8* src_ptr, int, static void ScaleARGBRowDown2_C(const uint8* src_ptr,
ptrdiff_t /* src_stride */,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
const uint32* src = reinterpret_cast<const uint32*>(src_ptr); const uint32* src = reinterpret_cast<const uint32*>(src_ptr);
uint32* dst = reinterpret_cast<uint32*>(dst_ptr); uint32* dst = reinterpret_cast<uint32*>(dst_ptr);
@ -659,7 +662,7 @@ static void ScaleARGBRowDown2_C(const uint8* src_ptr, int,
} }
} }
static void ScaleARGBRowDown2Int_C(const uint8* src_ptr, int src_stride, static void ScaleARGBRowDown2Int_C(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
for (int x = 0; x < dst_width; ++x) { for (int x = 0; x < dst_width; ++x) {
dst_ptr[0] = (src_ptr[0] + src_ptr[4] + dst_ptr[0] = (src_ptr[0] + src_ptr[4] +
@ -675,7 +678,7 @@ static void ScaleARGBRowDown2Int_C(const uint8* src_ptr, int src_stride,
} }
} }
void ScaleARGBRowDownEven_C(const uint8* src_ptr, int, void ScaleARGBRowDownEven_C(const uint8* src_ptr, ptrdiff_t /* src_stride */,
int src_stepx, int src_stepx,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
const uint32* src = reinterpret_cast<const uint32*>(src_ptr); const uint32* src = reinterpret_cast<const uint32*>(src_ptr);
@ -692,7 +695,8 @@ void ScaleARGBRowDownEven_C(const uint8* src_ptr, int,
} }
} }
static void ScaleARGBRowDownEvenInt_C(const uint8* src_ptr, int src_stride, static void ScaleARGBRowDownEvenInt_C(const uint8* src_ptr,
ptrdiff_t src_stride,
int src_stepx, int src_stepx,
uint8* dst_ptr, int dst_width) { uint8* dst_ptr, int dst_width) {
for (int x = 0; x < dst_width; ++x) { for (int x = 0; x < dst_width; ++x) {
@ -749,7 +753,8 @@ static void ScaleARGBFilterCols_C(uint8* dst_ptr, const uint8* src_ptr,
static const int kMaxInputWidth = 2560; static const int kMaxInputWidth = 2560;
// C version 2x2 -> 2x1 // C version 2x2 -> 2x1
void ScaleARGBFilterRows_C(uint8* dst_ptr, const uint8* src_ptr, int src_stride, void ScaleARGBFilterRows_C(uint8* dst_ptr, const uint8* src_ptr,
ptrdiff_t src_stride,
int dst_width, int source_y_fraction) { int dst_width, int source_y_fraction) {
assert(dst_width > 0); assert(dst_width > 0);
int y1_fraction = source_y_fraction; int y1_fraction = source_y_fraction;
@ -790,8 +795,8 @@ static void ScaleARGBDown2(int src_width, int src_height,
FilterMode filtering) { FilterMode filtering) {
assert(IS_ALIGNED(src_width, 2)); assert(IS_ALIGNED(src_width, 2));
assert(IS_ALIGNED(src_height, 2)); assert(IS_ALIGNED(src_height, 2));
void (*ScaleARGBRowDown2)(const uint8* src_ptr, int src_stride, void (*ScaleARGBRowDown2)(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst_ptr, int dst_width) = uint8* dst_ptr, int dst_width) =
filtering ? ScaleARGBRowDown2Int_C : ScaleARGBRowDown2_C; filtering ? ScaleARGBRowDown2Int_C : ScaleARGBRowDown2_C;
#if defined(HAS_SCALEARGBROWDOWN2_SSE2) #if defined(HAS_SCALEARGBROWDOWN2_SSE2)
if (TestCpuFlag(kCpuHasSSE2) && if (TestCpuFlag(kCpuHasSSE2) &&
@ -825,7 +830,7 @@ static void ScaleARGBDownEven(int src_width, int src_height,
FilterMode filtering) { FilterMode filtering) {
assert(IS_ALIGNED(src_width, 2)); assert(IS_ALIGNED(src_width, 2));
assert(IS_ALIGNED(src_height, 2)); assert(IS_ALIGNED(src_height, 2));
void (*ScaleARGBRowDownEven)(const uint8* src_ptr, int src_stride, void (*ScaleARGBRowDownEven)(const uint8* src_ptr, ptrdiff_t src_stride,
int src_step, uint8* dst_ptr, int dst_width) = int src_step, uint8* dst_ptr, int dst_width) =
filtering ? ScaleARGBRowDownEvenInt_C : ScaleARGBRowDownEven_C; filtering ? ScaleARGBRowDownEvenInt_C : ScaleARGBRowDownEven_C;
#if defined(HAS_SCALEARGBROWDOWNEVEN_SSE2) #if defined(HAS_SCALEARGBROWDOWNEVEN_SSE2)
@ -861,7 +866,7 @@ static void ScaleARGBBilinear(int src_width, int src_height,
assert(src_width <= kMaxInputWidth); assert(src_width <= kMaxInputWidth);
SIMD_ALIGNED(uint8 row[kMaxInputWidth * 4 + 16]); SIMD_ALIGNED(uint8 row[kMaxInputWidth * 4 + 16]);
void (*ScaleARGBFilterRows)(uint8* dst_ptr, const uint8* src_ptr, void (*ScaleARGBFilterRows)(uint8* dst_ptr, const uint8* src_ptr,
int src_stride, ptrdiff_t src_stride,
int dst_width, int source_y_fraction) = int dst_width, int source_y_fraction) =
ScaleARGBFilterRows_C; ScaleARGBFilterRows_C;
#if defined(HAS_SCALEARGBFILTERROWS_SSE2) #if defined(HAS_SCALEARGBFILTERROWS_SSE2)