coffee/libyuv
1
0
Fork 0
mirror of https://chromium.googlesource.com/libyuv/libyuv synced 2025-12-07 17:26:49 +08:00
Code Issues Packages Projects Releases Wiki Activity

Optimize the following 19 functions with LSX in row_lsx.cc.

Browse Source 
UYVYToYRow_LSX, UYVYToUVRow_LSX, UYVYToUV422Row_LSX,
ARGBToUVRow_LSX, ARGBToRGB24Row_LSX, ARGBToRAWRow_LSX,
ARGBToRGB565Row_LSX, ARGBToARGB1555Row_LSX, ARGBToARGB4444Row_LSX,
ARGBToUV444Row_LSX, ARGBMultiplyRow_LSX, ARGBAddRow_LSX,
ARGBSubtractRow_LSX, ARGBAttenuateRow_LSX, ARGBToRGB565DitherRow_LSX,
ARGBShuffleRow_LSX, ARGBShadeRow_LSX, ARGBGrayRow_LSX,
ARGBSepiaRow_LSX

Bug: libyuv:913
Change-Id: I02c0c9d68b229c4a66c96837e9b928c2f5dda1f3
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/4546814
Reviewed-by: Frank Barchard <fbarchard@chromium.org>
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
This commit is contained in:
Lu Wang 2023-05-19 16:46:54 +08:00 committed by libyuv LUCI CQ
parent a37799344d
commit 8670bcf17f
7 changed files with 968 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

135
include/libyuv/row.h
View File

@ -650,14 +650,30 @@ extern "C" {
#define HAS_ARGB1555TOUVROW_LSX
#define HAS_ARGB1555TOYROW_LSX
#define HAS_ARGB4444TOARGBROW_LSX
#define HAS_ARGBADDROW_LSX
#define HAS_ARGBATTENUATEROW_LSX
#define HAS_ARGBBLENDROW_LSX
#define HAS_ARGBCOLORMATRIXROW_LSX
#define HAS_ARGBEXTRACTALPHAROW_LSX
#define HAS_ARGBGRAYROW_LSX
#define HAS_ARGBSEPIAROW_LSX
#define HAS_ARGBSHADEROW_LSX
#define HAS_ARGBSHUFFLEROW_LSX
#define HAS_ARGBSUBTRACTROW_LSX
#define HAS_ARGBQUANTIZEROW_LSX
#define HAS_ARGBSETROW_LSX
#define HAS_ARGBTOARGB1555ROW_LSX
#define HAS_ARGBTOARGB4444ROW_LSX
#define HAS_ARGBTORAWROW_LSX
#define HAS_ARGBTORGB24ROW_LSX
#define HAS_ARGBTORGB565ROW_LSX
#define HAS_ARGBTORGB565DITHERROW_LSX
#define HAS_ARGBTOUVJROW_LSX
#define HAS_ARGBTOUV444ROW_LSX
#define HAS_ARGBTOUVROW_LSX
#define HAS_ARGBTOYJROW_LSX
#define HAS_ARGBMIRRORROW_LSX
#define HAS_ARGBMULTIPLYROW_LSX
#define HAS_BGRATOUVROW_LSX
#define HAS_BGRATOYROW_LSX
#define HAS_I400TOARGBROW_LSX
@ -697,6 +713,9 @@ extern "C" {
#define HAS_SOBELXYROW_LSX
#define HAS_SPLITUVROW_LSX
#define HAS_UYVYTOARGBROW_LSX
#define HAS_UYVYTOUV422ROW_LSX
#define HAS_UYVYTOUVROW_LSX
#define HAS_UYVYTOYROW_LSX
#define HAS_YUY2TOARGBROW_LSX
#define HAS_YUY2TOUVROW_LSX
#define HAS_YUY2TOUV422ROW_LSX
@ -798,6 +817,9 @@ extern "C" {
#define HAS_MERGEARGBROW_RVV
#define HAS_MERGERGBROW_RVV
#define HAS_MERGEXRGBROW_RVV
#define HAS_SPLITARGBROW_RVV
#define HAS_SPLITRGBROW_RVV
#define HAS_SPLITXRGBROW_RVV
#define HAS_RAWTOARGBROW_RVV
#define HAS_RAWTORGB24ROW_RVV
#define HAS_RAWTORGBAROW_RVV
@ -1387,11 +1409,20 @@ void ARGBToUVRow_MSA(const uint8_t* src_argb,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void ARGBToUVRow_LSX(const uint8_t* src_argb,
int src_stride_argb,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void ARGBToUVRow_LASX(const uint8_t* src_argb,
int src_stride_argb,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void ARGBToUV444Row_LSX(const uint8_t* src_argb,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void ARGBToUV444Row_LASX(const uint8_t* src_argb,
uint8_t* dst_u,
uint8_t* dst_v,
@ -1825,11 +1856,20 @@ void ARGBToUVRow_Any_MSA(const uint8_t* src_ptr,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void ARGBToUVRow_Any_LSX(const uint8_t* src_ptr,
int src_stride_ptr,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void ARGBToUVRow_Any_LASX(const uint8_t* src_ptr,
int src_stride_ptr,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void ARGBToUV444Row_Any_LSX(const uint8_t* src_ptr,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void ARGBToUV444Row_Any_LASX(const uint8_t* src_ptr,
uint8_t* dst_u,
uint8_t* dst_v,
@ -3095,6 +3135,10 @@ void ARGBShuffleRow_MSA(const uint8_t* src_argb,
uint8_t* dst_argb,
const uint8_t* shuffler,
int width);
void ARGBShuffleRow_LSX(const uint8_t* src_argb,
uint8_t* dst_argb,
const uint8_t* shuffler,
int width);
void ARGBShuffleRow_LASX(const uint8_t* src_argb,
uint8_t* dst_argb,
const uint8_t* shuffler,
@ -3115,6 +3159,10 @@ void ARGBShuffleRow_Any_MSA(const uint8_t* src_ptr,
uint8_t* dst_ptr,
const uint8_t* param,
int width);
void ARGBShuffleRow_Any_LSX(const uint8_t* src_ptr,
uint8_t* dst_ptr,
const uint8_t* param,
int width);
void ARGBShuffleRow_Any_LASX(const uint8_t* src_ptr,
uint8_t* dst_ptr,
const uint8_t* param,
@ -3369,17 +3417,30 @@ void ARGBToRGB565DitherRow_MSA(const uint8_t* src_argb,
uint8_t* dst_rgb,
const uint32_t dither4,
int width);
void ARGBToRGB565DitherRow_LSX(const uint8_t* src_argb,
uint8_t* dst_rgb,
const uint32_t dither4,
int width);
void ARGBToRGB565DitherRow_LASX(const uint8_t* src_argb,
uint8_t* dst_rgb,
const uint32_t dither4,
int width);
void ARGBToRGB24Row_LSX(const uint8_t* src_argb, uint8_t* dst_rgb, int width);
void ARGBToRGB24Row_LASX(const uint8_t* src_argb, uint8_t* dst_rgb, int width);
void ARGBToRAWRow_LSX(const uint8_t* src_argb, uint8_t* dst_rgb, int width);
void ARGBToRAWRow_LASX(const uint8_t* src_argb, uint8_t* dst_rgb, int width);
void ARGBToRGB565Row_LSX(const uint8_t* src_argb, uint8_t* dst_rgb, int width);
void ARGBToRGB565Row_LASX(const uint8_t* src_argb, uint8_t* dst_rgb, int width);
void ARGBToARGB1555Row_LSX(const uint8_t* src_argb,
uint8_t* dst_rgb,
int width);
void ARGBToARGB1555Row_LASX(const uint8_t* src_argb,
uint8_t* dst_rgb,
int width);
void ARGBToARGB4444Row_LSX(const uint8_t* src_argb,
uint8_t* dst_rgb,
int width);
void ARGBToARGB4444Row_LASX(const uint8_t* src_argb,
uint8_t* dst_rgb,
int width);
@ -4507,10 +4568,18 @@ void ARGBMultiplyRow_Any_MSA(const uint8_t* y_buf,
const uint8_t* uv_buf,
uint8_t* dst_ptr,
int width);
void ARGBMultiplyRow_LSX(const uint8_t* src_argb0,
const uint8_t* src_argb1,
uint8_t* dst_argb,
int width);
void ARGBMultiplyRow_LASX(const uint8_t* src_argb0,
const uint8_t* src_argb1,
uint8_t* dst_argb,
int width);
void ARGBMultiplyRow_Any_LSX(const uint8_t* y_buf,
const uint8_t* uv_buf,
uint8_t* dst_ptr,
int width);
void ARGBMultiplyRow_Any_LASX(const uint8_t* y_buf,
const uint8_t* uv_buf,
uint8_t* dst_ptr,
@ -4553,10 +4622,18 @@ void ARGBAddRow_Any_MSA(const uint8_t* y_buf,
const uint8_t* uv_buf,
uint8_t* dst_ptr,
int width);
void ARGBAddRow_LSX(const uint8_t* src_argb0,
const uint8_t* src_argb1,
uint8_t* dst_argb,
int width);
void ARGBAddRow_LASX(const uint8_t* src_argb0,
const uint8_t* src_argb1,
uint8_t* dst_argb,
int width);
void ARGBAddRow_Any_LSX(const uint8_t* y_buf,
const uint8_t* uv_buf,
uint8_t* dst_ptr,
int width);
void ARGBAddRow_Any_LASX(const uint8_t* y_buf,
const uint8_t* uv_buf,
uint8_t* dst_ptr,
@ -4600,10 +4677,18 @@ void ARGBSubtractRow_Any_MSA(const uint8_t* y_buf,
const uint8_t* uv_buf,
uint8_t* dst_ptr,
int width);
void ARGBSubtractRow_LSX(const uint8_t* src_argb0,
const uint8_t* src_argb1,
uint8_t* dst_argb,
int width);
void ARGBSubtractRow_LASX(const uint8_t* src_argb0,
const uint8_t* src_argb1,
uint8_t* dst_argb,
int width);
void ARGBSubtractRow_Any_LSX(const uint8_t* y_buf,
const uint8_t* uv_buf,
uint8_t* dst_ptr,
int width);
void ARGBSubtractRow_Any_LASX(const uint8_t* y_buf,
const uint8_t* uv_buf,
uint8_t* dst_ptr,
@ -4696,21 +4781,37 @@ void ARGBToRGB565DitherRow_Any_MSA(const uint8_t* src_ptr,
uint8_t* dst_ptr,
const uint32_t param,
int width);
void ARGBToRGB565DitherRow_Any_LSX(const uint8_t* src_ptr,
uint8_t* dst_ptr,
const uint32_t param,
int width);
void ARGBToRGB565DitherRow_Any_LASX(const uint8_t* src_ptr,
uint8_t* dst_ptr,
const uint32_t param,
int width);
void ARGBToRGB24Row_Any_LSX(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int width);
void ARGBToRGB24Row_Any_LASX(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int width);
void ARGBToRAWRow_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
void ARGBToRAWRow_Any_LASX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
void ARGBToRGB565Row_Any_LSX(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int width);
void ARGBToRGB565Row_Any_LASX(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int width);
void ARGBToARGB1555Row_Any_LSX(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int width);
void ARGBToARGB1555Row_Any_LASX(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int width);
void ARGBToARGB4444Row_Any_LSX(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int width);
void ARGBToARGB4444Row_Any_LASX(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int width);
@ -5257,12 +5358,18 @@ void UYVYToUV422Row_NEON(const uint8_t* src_uyvy,
uint8_t* dst_v,
int width);
void UYVYToYRow_MSA(const uint8_t* src_uyvy, uint8_t* dst_y, int width);
void UYVYToYRow_LSX(const uint8_t* src_uyvy, uint8_t* dst_y, int width);
void UYVYToYRow_LASX(const uint8_t* src_uyvy, uint8_t* dst_y, int width);
void UYVYToUVRow_MSA(const uint8_t* src_uyvy,
int src_stride_uyvy,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void UYVYToUVRow_LSX(const uint8_t* src_uyvy,
int src_stride_uyvy,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void UYVYToUVRow_LASX(const uint8_t* src_uyvy,
int src_stride_uyvy,
uint8_t* dst_u,
@ -5272,6 +5379,10 @@ void UYVYToUV422Row_MSA(const uint8_t* src_uyvy,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void UYVYToUV422Row_LSX(const uint8_t* src_uyvy,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void UYVYToUV422Row_LASX(const uint8_t* src_uyvy,
uint8_t* dst_u,
uint8_t* dst_v,
@ -5318,12 +5429,18 @@ void UYVYToUV422Row_Any_NEON(const uint8_t* src_ptr,
uint8_t* dst_v,
int width);
void UYVYToYRow_Any_MSA(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
void UYVYToYRow_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
void UYVYToYRow_Any_LASX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
void UYVYToUVRow_Any_MSA(const uint8_t* src_ptr,
int src_stride_ptr,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void UYVYToUVRow_Any_LSX(const uint8_t* src_ptr,
int src_stride_ptr,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void UYVYToUVRow_Any_LASX(const uint8_t* src_ptr,
int src_stride_ptr,
uint8_t* dst_u,
@ -5333,6 +5450,10 @@ void UYVYToUV422Row_Any_MSA(const uint8_t* src_ptr,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void UYVYToUV422Row_Any_LSX(const uint8_t* src_ptr,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void UYVYToUV422Row_Any_LASX(const uint8_t* src_ptr,
uint8_t* dst_u,
uint8_t* dst_v,
@ -5517,6 +5638,9 @@ void ARGBAttenuateRow_NEON(const uint8_t* src_argb,
void ARGBAttenuateRow_MSA(const uint8_t* src_argb,
uint8_t* dst_argb,
int width);
void ARGBAttenuateRow_LSX(const uint8_t* src_argb,
uint8_t* dst_argb,
int width);
void ARGBAttenuateRow_LASX(const uint8_t* src_argb,
uint8_t* dst_argb,
int width);
@ -5535,6 +5659,9 @@ void ARGBAttenuateRow_Any_NEON(const uint8_t* src_ptr,
void ARGBAttenuateRow_Any_MSA(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int width);
void ARGBAttenuateRow_Any_LSX(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int width);
void ARGBAttenuateRow_Any_LASX(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int width);
@ -5561,12 +5688,14 @@ void ARGBGrayRow_C(const uint8_t* src_argb, uint8_t* dst_argb, int width);
void ARGBGrayRow_SSSE3(const uint8_t* src_argb, uint8_t* dst_argb, int width);
void ARGBGrayRow_NEON(const uint8_t* src_argb, uint8_t* dst_argb, int width);
void ARGBGrayRow_MSA(const uint8_t* src_argb, uint8_t* dst_argb, int width);
void ARGBGrayRow_LSX(const uint8_t* src_argb, uint8_t* dst_argb, int width);
void ARGBGrayRow_LASX(const uint8_t* src_argb, uint8_t* dst_argb, int width);
void ARGBSepiaRow_C(uint8_t* dst_argb, int width);
void ARGBSepiaRow_SSSE3(uint8_t* dst_argb, int width);
void ARGBSepiaRow_NEON(uint8_t* dst_argb, int width);
void ARGBSepiaRow_MSA(uint8_t* dst_argb, int width);
void ARGBSepiaRow_LSX(uint8_t* dst_argb, int width);
void ARGBSepiaRow_LASX(uint8_t* dst_argb, int width);
void ARGBColorMatrixRow_C(const uint8_t* src_argb,
@ -5646,6 +5775,10 @@ void ARGBShadeRow_MSA(const uint8_t* src_argb,
uint8_t* dst_argb,
int width,
uint32_t value);
void ARGBShadeRow_LSX(const uint8_t* src_argb,
uint8_t* dst_argb,
int width,
uint32_t value);
void ARGBShadeRow_LASX(const uint8_t* src_argb,
uint8_t* dst_argb,
int width,

40
source/convert.cc
View File

@ -1558,6 +1558,26 @@ int UYVYToI420(const uint8_t* src_uyvy,
}
}
#endif
#if defined(HAS_UYVYTOYROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
UYVYToYRow = UYVYToYRow_Any_LSX;
UYVYToUVRow = UYVYToUVRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
UYVYToYRow = UYVYToYRow_LSX;
UYVYToUVRow = UYVYToUVRow_LSX;
}
}
#endif
#if defined(HAS_UYVYTOYROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
UYVYToYRow = UYVYToYRow_Any_LSX;
UYVYToUVRow = UYVYToUVRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
UYVYToYRow = UYVYToYRow_LSX;
UYVYToUVRow = UYVYToUVRow_LSX;
}
}
#endif
#if defined(HAS_UYVYTOYROW_LASX)
if (TestCpuFlag(kCpuHasLASX)) {
UYVYToYRow = UYVYToYRow_Any_LASX;
@ -1818,6 +1838,16 @@ int ARGBToI420(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTOYROW_LSX) && defined(HAS_ARGBTOUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToYRow = ARGBToYRow_Any_LSX;
ARGBToUVRow = ARGBToUVRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
ARGBToYRow = ARGBToYRow_LSX;
ARGBToUVRow = ARGBToUVRow_LSX;
}
}
#endif
#if defined(HAS_ARGBTOYROW_LASX) && defined(HAS_ARGBTOUVROW_LASX)
if (TestCpuFlag(kCpuHasLASX)) {
ARGBToYRow = ARGBToYRow_Any_LASX;
@ -3610,6 +3640,16 @@ int ARGB4444ToI420(const uint8_t* src_argb4444,
}
}
#endif
#if defined(HAS_ARGBTOYROW_LSX) && defined(HAS_ARGBTOUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToYRow = ARGBToYRow_Any_LSX;
ARGBToUVRow = ARGBToUVRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
ARGBToYRow = ARGBToYRow_LSX;
ARGBToUVRow = ARGBToUVRow_LSX;
}
}
#endif
#if defined(HAS_ARGBTOYROW_LASX) && defined(HAS_ARGBTOUVROW_LASX)
if (TestCpuFlag(kCpuHasLASX)) {
ARGBToYRow = ARGBToYRow_Any_LASX;

8
source/convert_argb.cc
View File

@ -5742,6 +5742,14 @@ int I420ToRGB565Dither(const uint8_t* src_y,
}
}
#endif
#if defined(HAS_ARGBTORGB565DITHERROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_Any_LSX;
if (IS_ALIGNED(width, 8)) {
ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_LSX;
}
}
#endif
#if defined(HAS_ARGBTORGB565DITHERROW_LASX)
if (TestCpuFlag(kCpuHasLASX)) {
ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_Any_LASX;

97
source/convert_from_argb.cc
View File

@ -76,6 +76,14 @@ int ARGBToI444(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTOUV444ROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToUV444Row = ARGBToUV444Row_Any_LSX;
if (IS_ALIGNED(width, 16)) {
ARGBToUV444Row = ARGBToUV444Row_LSX;
}
}
#endif
#if defined(HAS_ARGBTOUV444ROW_LASX)
if (TestCpuFlag(kCpuHasLASX)) {
ARGBToUV444Row = ARGBToUV444Row_Any_LASX;
@ -251,6 +259,16 @@ int ARGBToI422(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTOYROW_LSX) && defined(HAS_ARGBTOUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToYRow = ARGBToYRow_Any_LSX;
ARGBToUVRow = ARGBToUVRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
ARGBToYRow = ARGBToYRow_LSX;
ARGBToUVRow = ARGBToUVRow_LSX;
}
}
#endif
#if defined(HAS_ARGBTOYROW_LASX) && defined(HAS_ARGBTOUVROW_LASX)
if (TestCpuFlag(kCpuHasLASX)) {
ARGBToYRow = ARGBToYRow_Any_LASX;
@ -556,6 +574,16 @@ int ARGBToNV21(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTOYROW_LSX) && defined(HAS_ARGBTOUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToYRow = ARGBToYRow_Any_LSX;
ARGBToUVRow = ARGBToUVRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
ARGBToYRow = ARGBToYRow_LSX;
ARGBToUVRow = ARGBToUVRow_LSX;
}
}
#endif
#if defined(HAS_ARGBTOYROW_LASX) && defined(HAS_ARGBTOUVROW_LASX)
if (TestCpuFlag(kCpuHasLASX)) {
ARGBToYRow = ARGBToYRow_Any_LASX;
@ -1105,6 +1133,16 @@ int ARGBToYUY2(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTOYROW_LSX) && defined(HAS_ARGBTOUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToYRow = ARGBToYRow_Any_LSX;
ARGBToUVRow = ARGBToUVRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
ARGBToYRow = ARGBToYRow_LSX;
ARGBToUVRow = ARGBToUVRow_LSX;
}
}
#endif
#if defined(HAS_ARGBTOYROW_LASX) && defined(HAS_ARGBTOUVROW_LASX)
if (TestCpuFlag(kCpuHasLASX)) {
ARGBToYRow = ARGBToYRow_Any_LASX;
@ -1289,6 +1327,16 @@ int ARGBToUYVY(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTOYROW_LSX) && defined(HAS_ARGBTOUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToYRow = ARGBToYRow_Any_LSX;
ARGBToUVRow = ARGBToUVRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
ARGBToYRow = ARGBToYRow_LSX;
ARGBToUVRow = ARGBToUVRow_LSX;
}
}
#endif
#if defined(HAS_ARGBTOYROW_LASX) && defined(HAS_ARGBTOUVROW_LASX)
if (TestCpuFlag(kCpuHasLASX)) {
ARGBToYRow = ARGBToYRow_Any_LASX;
@ -1540,6 +1588,14 @@ int ARGBToRGB24(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTORGB24ROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToRGB24Row = ARGBToRGB24Row_Any_LSX;
if (IS_ALIGNED(width, 16)) {
ARGBToRGB24Row = ARGBToRGB24Row_LSX;
}
}
#endif
#if defined(HAS_ARGBTORGB24ROW_LASX)
if (TestCpuFlag(kCpuHasLASX)) {
ARGBToRGB24Row = ARGBToRGB24Row_Any_LASX;
@ -1619,6 +1675,14 @@ int ARGBToRAW(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTORAWROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToRAWRow = ARGBToRAWRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
ARGBToRAWRow = ARGBToRAWRow_LSX;
}
}
#endif
#if defined(HAS_ARGBTORAWROW_LASX)
if (TestCpuFlag(kCpuHasLASX)) {
ARGBToRAWRow = ARGBToRAWRow_Any_LASX;
@ -1702,6 +1766,14 @@ int ARGBToRGB565Dither(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTORGB565DITHERROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_Any_LSX;
if (IS_ALIGNED(width, 8)) {
ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_LSX;
}
}
#endif
#if defined(HAS_ARGBTORGB565DITHERROW_LASX)
if (TestCpuFlag(kCpuHasLASX)) {
ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_Any_LASX;
@ -1779,6 +1851,15 @@ int ARGBToRGB565(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTORGB565ROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToRGB565Row = ARGBToRGB565Row_Any_LSX;
if (IS_ALIGNED(width, 8)) {
ARGBToRGB565Row = ARGBToRGB565Row_LSX;
}
}
#endif
#if defined(HAS_ARGBTORGB565ROW_LASX)
if (TestCpuFlag(kCpuHasLASX)) {
ARGBToRGB565Row = ARGBToRGB565Row_Any_LASX;
@ -1853,6 +1934,14 @@ int ARGBToARGB1555(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTOARGB1555ROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToARGB1555Row = ARGBToARGB1555Row_Any_LSX;
if (IS_ALIGNED(width, 8)) {
ARGBToARGB1555Row = ARGBToARGB1555Row_LSX;
}
}
#endif
#if defined(HAS_ARGBTOARGB1555ROW_LASX)
if (TestCpuFlag(kCpuHasLASX)) {
ARGBToARGB1555Row = ARGBToARGB1555Row_Any_LASX;
@ -1927,6 +2016,14 @@ int ARGBToARGB4444(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTOARGB4444ROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToARGB4444Row = ARGBToARGB4444Row_Any_LSX;
if (IS_ALIGNED(width, 8)) {
ARGBToARGB4444Row = ARGBToARGB4444Row_LSX;
}
}
#endif
#if defined(HAS_ARGBTOARGB4444ROW_LASX)
if (TestCpuFlag(kCpuHasLASX)) {
ARGBToARGB4444Row = ARGBToARGB4444Row_Any_LASX;

78
source/planar_functions.cc
View File

@ -2219,6 +2219,16 @@ int UYVYToI422(const uint8_t* src_uyvy,
}
}
#endif
#if defined(HAS_UYVYTOYROW_LSX) && defined(HAS_UYVYTOUV422ROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
UYVYToYRow = UYVYToYRow_Any_LSX;
UYVYToUV422Row = UYVYToUV422Row_Any_LSX;
if (IS_ALIGNED(width, 16)) {
UYVYToYRow = UYVYToYRow_LSX;
UYVYToUV422Row = UYVYToUV422Row_LSX;
}
}
#endif
#if defined(HAS_UYVYTOYROW_LASX) && defined(HAS_UYVYTOUV422ROW_LASX)
if (TestCpuFlag(kCpuHasLASX)) {
UYVYToYRow = UYVYToYRow_Any_LASX;
@ -2366,6 +2376,14 @@ int UYVYToY(const uint8_t* src_uyvy,
}
}
#endif
#if defined(HAS_UYVYTOYROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
UYVYToYRow = UYVYToYRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
UYVYToYRow = UYVYToYRow_LSX;
}
}
#endif
for (y = 0; y < height; ++y) {
UYVYToYRow(src_uyvy, dst_y, width);
@ -3068,6 +3086,14 @@ int ARGBMultiply(const uint8_t* src_argb0,
}
}
#endif
#if defined(HAS_ARGBMULTIPLYROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBMultiplyRow = ARGBMultiplyRow_Any_LSX;
if (IS_ALIGNED(width, 4)) {
ARGBMultiplyRow = ARGBMultiplyRow_LSX;
}
}
#endif
#if defined(HAS_ARGBMULTIPLYROW_LASX)
if (TestCpuFlag(kCpuHasLASX)) {
ARGBMultiplyRow = ARGBMultiplyRow_Any_LASX;
@ -3153,6 +3179,14 @@ int ARGBAdd(const uint8_t* src_argb0,
}
}
#endif
#if defined(HAS_ARGBADDROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBAddRow = ARGBAddRow_Any_LSX;
if (IS_ALIGNED(width, 4)) {
ARGBAddRow = ARGBAddRow_LSX;
}
}
#endif
#if defined(HAS_ARGBADDROW_LASX)
if (TestCpuFlag(kCpuHasLASX)) {
ARGBAddRow = ARGBAddRow_Any_LASX;
@ -3233,6 +3267,14 @@ int ARGBSubtract(const uint8_t* src_argb0,
}
}
#endif
#if defined(HAS_ARGBSUBTRACTROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBSubtractRow = ARGBSubtractRow_Any_LSX;
if (IS_ALIGNED(width, 4)) {
ARGBSubtractRow = ARGBSubtractRow_LSX;
}
}
#endif
#if defined(HAS_ARGBSUBTRACTROW_LASX)
if (TestCpuFlag(kCpuHasLASX)) {
ARGBSubtractRow = ARGBSubtractRow_Any_LASX;
@ -3558,6 +3600,14 @@ int ARGBAttenuate(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBAttenuateRow = ARGBAttenuateRow_Any_LSX;
if (IS_ALIGNED(width, 8)) {
ARGBAttenuateRow = ARGBAttenuateRow_LSX;
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_LASX)
if (TestCpuFlag(kCpuHasLASX)) {
ARGBAttenuateRow = ARGBAttenuateRow_Any_LASX;
@ -3671,6 +3721,11 @@ int ARGBGrayTo(const uint8_t* src_argb,
ARGBGrayRow = ARGBGrayRow_MSA;
}
#endif
#if defined(HAS_ARGBGRAYROW_LSX)
if (TestCpuFlag(kCpuHasLSX) && IS_ALIGNED(width, 8)) {
ARGBGrayRow = ARGBGrayRow_LSX;
}
#endif
#if defined(HAS_ARGBGRAYROW_LASX)
if (TestCpuFlag(kCpuHasLASX) && IS_ALIGNED(width, 16)) {
ARGBGrayRow = ARGBGrayRow_LASX;
@ -3721,6 +3776,11 @@ int ARGBGray(uint8_t* dst_argb,
ARGBGrayRow = ARGBGrayRow_MSA;
}
#endif
#if defined(HAS_ARGBGRAYROW_LSX)
if (TestCpuFlag(kCpuHasLSX) && IS_ALIGNED(width, 8)) {
ARGBGrayRow = ARGBGrayRow_LSX;
}
#endif
#if defined(HAS_ARGBGRAYROW_LASX)
if (TestCpuFlag(kCpuHasLASX) && IS_ALIGNED(width, 16)) {
ARGBGrayRow = ARGBGrayRow_LASX;
@ -3769,6 +3829,11 @@ int ARGBSepia(uint8_t* dst_argb,
ARGBSepiaRow = ARGBSepiaRow_MSA;
}
#endif
#if defined(HAS_ARGBSEPIAROW_LSX)
if (TestCpuFlag(kCpuHasLSX) && IS_ALIGNED(width, 8)) {
ARGBSepiaRow = ARGBSepiaRow_LSX;
}
#endif
#if defined(HAS_ARGBSEPIAROW_LASX)
if (TestCpuFlag(kCpuHasLASX) && IS_ALIGNED(width, 16)) {
ARGBSepiaRow = ARGBSepiaRow_LASX;
@ -4194,6 +4259,11 @@ int ARGBShade(const uint8_t* src_argb,
ARGBShadeRow = ARGBShadeRow_MSA;
}
#endif
#if defined(HAS_ARGBSHADEROW_LSX)
if (TestCpuFlag(kCpuHasLSX) && IS_ALIGNED(width, 4)) {
ARGBShadeRow = ARGBShadeRow_LSX;
}
#endif
#if defined(HAS_ARGBSHADEROW_LASX)
if (TestCpuFlag(kCpuHasLASX) && IS_ALIGNED(width, 8)) {
ARGBShadeRow = ARGBShadeRow_LASX;
@ -4483,6 +4553,14 @@ int ARGBShuffle(const uint8_t* src_bgra,
}
}
#endif
#if defined(HAS_ARGBSHUFFLEROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBShuffleRow = ARGBShuffleRow_Any_LSX;
if (IS_ALIGNED(width, 8)) {
ARGBShuffleRow = ARGBShuffleRow_LSX;
}
}
#endif
#if defined(HAS_ARGBSHUFFLEROW_LASX)
if (TestCpuFlag(kCpuHasLASX)) {
ARGBShuffleRow = ARGBShuffleRow_Any_LASX;

41
source/row_any.cc
View File

@ -639,18 +639,27 @@ ANY21(ARGBSubtractRow_Any_NEON, ARGBSubtractRow_NEON, 0, 4, 4, 4, 7)
#ifdef HAS_ARGBMULTIPLYROW_MSA
ANY21(ARGBMultiplyRow_Any_MSA, ARGBMultiplyRow_MSA, 0, 4, 4, 4, 3)
#endif
#ifdef HAS_ARGBMULTIPLYROW_LSX
ANY21(ARGBMultiplyRow_Any_LSX, ARGBMultiplyRow_LSX, 0, 4, 4, 4, 3)
#endif
#ifdef HAS_ARGBMULTIPLYROW_LASX
ANY21(ARGBMultiplyRow_Any_LASX, ARGBMultiplyRow_LASX, 0, 4, 4, 4, 7)
#endif
#ifdef HAS_ARGBADDROW_MSA
ANY21(ARGBAddRow_Any_MSA, ARGBAddRow_MSA, 0, 4, 4, 4, 7)
#endif
#ifdef HAS_ARGBADDROW_LSX
ANY21(ARGBAddRow_Any_LSX, ARGBAddRow_LSX, 0, 4, 4, 4, 3)
#endif
#ifdef HAS_ARGBADDROW_LASX
ANY21(ARGBAddRow_Any_LASX, ARGBAddRow_LASX, 0, 4, 4, 4, 7)
#endif
#ifdef HAS_ARGBSUBTRACTROW_MSA
ANY21(ARGBSubtractRow_Any_MSA, ARGBSubtractRow_MSA, 0, 4, 4, 4, 7)
#endif
#ifdef HAS_ARGBSUBTRACTROW_LSX
ANY21(ARGBSubtractRow_Any_LSX, ARGBSubtractRow_LSX, 0, 4, 4, 4, 3)
#endif
#ifdef HAS_ARGBSUBTRACTROW_LASX
ANY21(ARGBSubtractRow_Any_LASX, ARGBSubtractRow_LASX, 0, 4, 4, 4, 7)
#endif
@ -992,6 +1001,13 @@ ANY11(ARGBToARGB1555Row_Any_MSA, ARGBToARGB1555Row_MSA, 0, 4, 2, 7)
ANY11(ARGBToARGB4444Row_Any_MSA, ARGBToARGB4444Row_MSA, 0, 4, 2, 7)
ANY11(J400ToARGBRow_Any_MSA, J400ToARGBRow_MSA, 0, 1, 4, 15)
#endif
#if defined(HAS_ARGBTORGB24ROW_LSX)
ANY11(ARGBToRGB24Row_Any_LSX, ARGBToRGB24Row_LSX, 0, 4, 3, 15)
ANY11(ARGBToRAWRow_Any_LSX, ARGBToRAWRow_LSX, 0, 4, 3, 15)
ANY11(ARGBToRGB565Row_Any_LSX, ARGBToRGB565Row_LSX, 0, 4, 2, 7)
ANY11(ARGBToARGB1555Row_Any_LSX, ARGBToARGB1555Row_LSX, 0, 4, 2, 7)
ANY11(ARGBToARGB4444Row_Any_LSX, ARGBToARGB4444Row_LSX, 0, 4, 2, 7)
#endif
#if defined(HAS_ARGBTORGB24ROW_LASX)
ANY11(ARGBToRGB24Row_Any_LASX, ARGBToRGB24Row_LASX, 0, 4, 3, 31)
ANY11(ARGBToRAWRow_Any_LASX, ARGBToRAWRow_LASX, 0, 4, 3, 31)
@ -1230,6 +1246,9 @@ ANY11(YUY2ToYRow_Any_LASX, YUY2ToYRow_LASX, 1, 4, 1, 31)
#ifdef HAS_UYVYTOYROW_MSA
ANY11(UYVYToYRow_Any_MSA, UYVYToYRow_MSA, 1, 4, 1, 31)
#endif
#ifdef HAS_UYVYTOYROW_LSX
ANY11(UYVYToYRow_Any_LSX, UYVYToYRow_LSX, 1, 4, 1, 15)
#endif
#ifdef HAS_UYVYTOYROW_LASX
ANY11(UYVYToYRow_Any_LASX, UYVYToYRow_LASX, 1, 4, 1, 31)
#endif
@ -1326,6 +1345,9 @@ ANY11(ARGBAttenuateRow_Any_NEON, ARGBAttenuateRow_NEON, 0, 4, 4, 7)
#ifdef HAS_ARGBATTENUATEROW_MSA
ANY11(ARGBAttenuateRow_Any_MSA, ARGBAttenuateRow_MSA, 0, 4, 4, 7)
#endif
#ifdef HAS_ARGBATTENUATEROW_LSX
ANY11(ARGBAttenuateRow_Any_LSX, ARGBAttenuateRow_LSX, 0, 4, 4, 7)
#endif
#ifdef HAS_ARGBATTENUATEROW_LASX
ANY11(ARGBAttenuateRow_Any_LASX, ARGBAttenuateRow_LASX, 0, 4, 4, 15)
#endif
@ -1467,6 +1489,14 @@ ANY11P(ARGBToRGB565DitherRow_Any_MSA,
2,
7)
#endif
#if defined(HAS_ARGBTORGB565DITHERROW_LSX)
ANY11P(ARGBToRGB565DitherRow_Any_LSX,
ARGBToRGB565DitherRow_LSX,
const uint32_t,
4,
2,
7)
#endif
#if defined(HAS_ARGBTORGB565DITHERROW_LASX)
ANY11P(ARGBToRGB565DitherRow_Any_LASX,
ARGBToRGB565DitherRow_LASX,
@ -1487,6 +1517,9 @@ ANY11P(ARGBShuffleRow_Any_NEON, ARGBShuffleRow_NEON, const uint8_t*, 4, 4, 3)
#ifdef HAS_ARGBSHUFFLEROW_MSA
ANY11P(ARGBShuffleRow_Any_MSA, ARGBShuffleRow_MSA, const uint8_t*, 4, 4, 7)
#endif
#ifdef HAS_ARGBSHUFFLEROW_LSX
ANY11P(ARGBShuffleRow_Any_LSX, ARGBShuffleRow_LSX, const uint8_t*, 4, 4, 7)
#endif
#ifdef HAS_ARGBSHUFFLEROW_LASX
ANY11P(ARGBShuffleRow_Any_LASX, ARGBShuffleRow_LASX, const uint8_t*, 4, 4, 15)
#endif
@ -2000,7 +2033,9 @@ ANY12(YUY2ToUV422Row_Any_MSA, YUY2ToUV422Row_MSA, 1, 4, 1, 31)
ANY12(UYVYToUV422Row_Any_MSA, UYVYToUV422Row_MSA, 1, 4, 1, 31)
#endif
#ifdef HAS_YUY2TOUV422ROW_LSX
ANY12(ARGBToUV444Row_Any_LSX, ARGBToUV444Row_LSX, 0, 4, 0, 15)
ANY12(YUY2ToUV422Row_Any_LSX, YUY2ToUV422Row_LSX, 1, 4, 1, 15)
ANY12(UYVYToUV422Row_Any_LSX, UYVYToUV422Row_LSX, 1, 4, 1, 15)
#endif
#ifdef HAS_YUY2TOUV422ROW_LASX
ANY12(ARGBToUV444Row_Any_LASX, ARGBToUV444Row_LASX, 0, 4, 0, 31)
@ -2172,6 +2207,9 @@ ANY12S(ARGBToUVRow_Any_NEON, ARGBToUVRow_NEON, 0, 4, 15)
#ifdef HAS_ARGBTOUVROW_MSA
ANY12S(ARGBToUVRow_Any_MSA, ARGBToUVRow_MSA, 0, 4, 31)
#endif
#ifdef HAS_ARGBTOUVROW_LSX
ANY12S(ARGBToUVRow_Any_LSX, ARGBToUVRow_LSX, 0, 4, 15)
#endif
#ifdef HAS_ARGBTOUVROW_LASX
ANY12S(ARGBToUVRow_Any_LASX, ARGBToUVRow_LASX, 0, 4, 31)
#endif
@ -2292,6 +2330,9 @@ ANY12S(YUY2ToUVRow_Any_LASX, YUY2ToUVRow_LASX, 1, 4, 31)
#ifdef HAS_UYVYTOUVROW_MSA
ANY12S(UYVYToUVRow_Any_MSA, UYVYToUVRow_MSA, 1, 4, 31)
#endif
#ifdef HAS_UYVYTOUVROW_LSX
ANY12S(UYVYToUVRow_Any_LSX, UYVYToUVRow_LSX, 1, 4, 15)
#endif
#ifdef HAS_UYVYTOUVROW_LASX
ANY12S(UYVYToUVRow_Any_LASX, UYVYToUVRow_LASX, 1, 4, 31)
#endif

570
source/row_lsx.cc
View File

@ -713,6 +713,576 @@ void YUY2ToUV422Row_LSX(const uint8_t* src_yuy2,
}
}
void UYVYToYRow_LSX(const uint8_t* src_uyvy, uint8_t* dst_y, int width) {
int x;
int len = width / 16;
__m128i src0, src1, dst0;
for (x = 0; x < len; x++) {
DUP2_ARG2(__lsx_vld, src_uyvy, 0, src_uyvy, 16, src0, src1);
dst0 = __lsx_vpickod_b(src1, src0);
__lsx_vst(dst0, dst_y, 0);
src_uyvy += 32;
dst_y += 16;
}
}
void UYVYToUVRow_LSX(const uint8_t* src_uyvy,
int src_stride_uyvy,
uint8_t* dst_u,
uint8_t* dst_v,
int width) {
const uint8_t* src_uyvy_next = src_uyvy + src_stride_uyvy;
int x;
int len = width / 16;
__m128i src0, src1, src2, src3, tmp0, dst0, dst1;
for (x = 0; x < len; x++) {
DUP4_ARG2(__lsx_vld, src_uyvy, 0, src_uyvy, 16, src_uyvy_next, 0,
src_uyvy_next, 16, src0, src1, src2, src3);
src0 = __lsx_vpickev_b(src1, src0);
src1 = __lsx_vpickev_b(src3, src2);
tmp0 = __lsx_vavgr_bu(src1, src0);
dst0 = __lsx_vpickev_b(tmp0, tmp0);
dst1 = __lsx_vpickod_b(tmp0, tmp0);
__lsx_vstelm_d(dst0, dst_u, 0, 0);
__lsx_vstelm_d(dst1, dst_v, 0, 0);
src_uyvy += 32;
src_uyvy_next += 32;
dst_u += 8;
dst_v += 8;
}
}
void UYVYToUV422Row_LSX(const uint8_t* src_uyvy,
uint8_t* dst_u,
uint8_t* dst_v,
int width) {
int x;
int len = width / 16;
__m128i src0, src1, tmp0, dst0, dst1;
for (x = 0; x < len; x++) {
DUP2_ARG2(__lsx_vld, src_uyvy, 0, src_uyvy, 16, src0, src1);
tmp0 = __lsx_vpickev_b(src1, src0);
dst0 = __lsx_vpickev_b(tmp0, tmp0);
dst1 = __lsx_vpickod_b(tmp0, tmp0);
__lsx_vstelm_d(dst0, dst_u, 0, 0);
__lsx_vstelm_d(dst1, dst_v, 0, 0);
src_uyvy += 32;
dst_u += 8;
dst_v += 8;
}
}
void ARGBToUVRow_LSX(const uint8_t* src_argb0,
int src_stride_argb,
uint8_t* dst_u,
uint8_t* dst_v,
int width) {
int x;
int len = width / 16;
const uint8_t* src_argb1 = src_argb0 + src_stride_argb;
__m128i src0, src1, src2, src3, src4, src5, src6, src7;
__m128i vec0, vec1, vec2, vec3;
__m128i tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, dst0, dst1;
__m128i const_0x70 = {0x0038003800380038, 0x0038003800380038};
__m128i const_0x4A = {0x0025002500250025, 0x0025002500250025};
__m128i const_0x26 = {0x0013001300130013, 0x0013001300130013};
__m128i const_0x5E = {0x002f002f002f002f, 0x002f002f002f002f};
__m128i const_0x12 = {0x0009000900090009, 0x0009000900090009};
__m128i const_0x8080 = {0x8080808080808080, 0x8080808080808080};
for (x = 0; x < len; x++) {
DUP4_ARG2(__lsx_vld, src_argb0, 0, src_argb0, 16, src_argb0, 32,
src_argb0, 48, src0, src1, src2, src3);
DUP4_ARG2(__lsx_vld, src_argb1, 0, src_argb1, 16, src_argb1, 32,
src_argb1, 48, src4, src5, src6, src7);
vec0 = __lsx_vaddwev_h_bu(src0, src4);
vec1 = __lsx_vaddwev_h_bu(src1, src5);
vec2 = __lsx_vaddwev_h_bu(src2, src6);
vec3 = __lsx_vaddwev_h_bu(src3, src7);
tmp0 = __lsx_vpickev_h(vec1, vec0);
tmp1 = __lsx_vpickev_h(vec3, vec2);
tmp2 = __lsx_vpickod_h(vec1, vec0);
tmp3 = __lsx_vpickod_h(vec3, vec2);
vec0 = __lsx_vaddwod_h_bu(src0, src4);
vec1 = __lsx_vaddwod_h_bu(src1, src5);
vec2 = __lsx_vaddwod_h_bu(src2, src6);
vec3 = __lsx_vaddwod_h_bu(src3, src7);
tmp4 = __lsx_vpickev_h(vec1, vec0);
tmp5 = __lsx_vpickev_h(vec3, vec2);
vec0 = __lsx_vpickev_h(tmp1, tmp0);
vec1 = __lsx_vpickod_h(tmp1, tmp0);
src0 = __lsx_vavgr_h(vec0, vec1);
vec0 = __lsx_vpickev_h(tmp3, tmp2);
vec1 = __lsx_vpickod_h(tmp3, tmp2);
src1 = __lsx_vavgr_h(vec0, vec1);
vec0 = __lsx_vpickev_h(tmp5, tmp4);
vec1 = __lsx_vpickod_h(tmp5, tmp4);
src2 = __lsx_vavgr_h(vec0, vec1);
dst0 = __lsx_vmadd_h(const_0x8080, src0, const_0x70);
dst0 = __lsx_vmsub_h(dst0, src2, const_0x4A);
dst0 = __lsx_vmsub_h(dst0, src1, const_0x26);
dst1 = __lsx_vmadd_h(const_0x8080, src1, const_0x70);
dst1 = __lsx_vmsub_h(dst1, src2, const_0x5E);
dst1 = __lsx_vmsub_h(dst1, src0, const_0x12);
dst0 = __lsx_vsrai_h(dst0, 8);
dst1 = __lsx_vsrai_h(dst1, 8);
dst0 = __lsx_vpickev_b(dst1, dst0);
__lsx_vstelm_d(dst0, dst_u, 0, 0);
__lsx_vstelm_d(dst0, dst_v, 0, 1);
src_argb0 += 64;
src_argb1 += 64;
dst_u += 8;
dst_v += 8;
}
}
void ARGBToRGB24Row_LSX(const uint8_t* src_argb, uint8_t* dst_rgb, int width) {
int x;
int len = (width / 16) - 1;
__m128i src0, src1, src2, src3;
__m128i tmp0, tmp1, tmp2, tmp3;
__m128i shuf = {0x0908060504020100, 0x000000000E0D0C0A};
for (x = 0; x < len; x++) {
DUP4_ARG2(__lsx_vld, src_argb, 0, src_argb, 16, src_argb, 32, src_argb,
48, src0, src1, src2, src3);
tmp0 = __lsx_vshuf_b(src0, src0, shuf);
tmp1 = __lsx_vshuf_b(src1, src1, shuf);
tmp2 = __lsx_vshuf_b(src2, src2, shuf);
tmp3 = __lsx_vshuf_b(src3, src3, shuf);
__lsx_vst(tmp0, dst_rgb, 0);
__lsx_vst(tmp1, dst_rgb, 12);
__lsx_vst(tmp2, dst_rgb, 24);
__lsx_vst(tmp3, dst_rgb, 36);
dst_rgb += 48;
src_argb += 64;
}
DUP4_ARG2(__lsx_vld, src_argb, 0, src_argb, 16, src_argb, 32, src_argb, 48,
src0, src1, src2, src3);
tmp0 = __lsx_vshuf_b(src0, src0, shuf);
tmp1 = __lsx_vshuf_b(src1, src1, shuf);
tmp2 = __lsx_vshuf_b(src2, src2, shuf);
tmp3 = __lsx_vshuf_b(src3, src3, shuf);
__lsx_vst(tmp0, dst_rgb, 0);
__lsx_vst(tmp1, dst_rgb, 12);
__lsx_vst(tmp2, dst_rgb, 24);
dst_rgb += 36;
__lsx_vst(tmp3, dst_rgb, 0);
}
void ARGBToRAWRow_LSX(const uint8_t* src_argb, uint8_t* dst_rgb, int width) {
int x;
int len = (width / 16) - 1;
__m128i src0, src1, src2, src3;
__m128i tmp0, tmp1, tmp2, tmp3;
__m128i shuf = {0x090A040506000102, 0x000000000C0D0E08};
for (x = 0; x < len; x++) {
DUP4_ARG2(__lsx_vld, src_argb, 0, src_argb, 16, src_argb, 32, src_argb,
48, src0, src1, src2, src3);
tmp0 = __lsx_vshuf_b(src0, src0, shuf);
tmp1 = __lsx_vshuf_b(src1, src1, shuf);
tmp2 = __lsx_vshuf_b(src2, src2, shuf);
tmp3 = __lsx_vshuf_b(src3, src3, shuf);
__lsx_vst(tmp0, dst_rgb, 0);
__lsx_vst(tmp1, dst_rgb, 12);
__lsx_vst(tmp2, dst_rgb, 24);
__lsx_vst(tmp3, dst_rgb, 36);
dst_rgb += 48;
src_argb += 64;
}
DUP4_ARG2(__lsx_vld, src_argb, 0, src_argb, 16, src_argb, 32, src_argb, 48,
src0, src1, src2, src3);
tmp0 = __lsx_vshuf_b(src0, src0, shuf);
tmp1 = __lsx_vshuf_b(src1, src1, shuf);
tmp2 = __lsx_vshuf_b(src2, src2, shuf);
tmp3 = __lsx_vshuf_b(src3, src3, shuf);
__lsx_vst(tmp0, dst_rgb, 0);
__lsx_vst(tmp1, dst_rgb, 12);
__lsx_vst(tmp2, dst_rgb, 24);
dst_rgb += 36;
__lsx_vst(tmp3, dst_rgb, 0);
}
void ARGBToRGB565Row_LSX(const uint8_t* src_argb,
uint8_t* dst_rgb,
int width) {
int x;
int len = width / 8;
__m128i zero = __lsx_vldi(0);
__m128i src0, src1, tmp0, tmp1, dst0;
__m128i shift = {0x0300030003000300, 0x0300030003000300};
for (x = 0; x < len; x++) {
DUP2_ARG2(__lsx_vld, src_argb, 0, src_argb, 16, src0, src1);
tmp0 = __lsx_vpickev_b(src1, src0);
tmp1 = __lsx_vpickod_b(src1, src0);
tmp0 = __lsx_vsrli_b(tmp0, 3);
tmp1 = __lsx_vpackev_b(zero, tmp1);
tmp1 = __lsx_vsrli_h(tmp1, 2);
tmp0 = __lsx_vsll_b(tmp0, shift);
tmp1 = __lsx_vslli_h(tmp1, 5);
dst0 = __lsx_vor_v(tmp0, tmp1);
__lsx_vst(dst0, dst_rgb, 0);
dst_rgb += 16;
src_argb += 32;
}
}
void ARGBToARGB1555Row_LSX(const uint8_t* src_argb,
uint8_t* dst_rgb,
int width) {
int x;
int len = width / 8;
__m128i zero = __lsx_vldi(0);
__m128i src0, src1, tmp0, tmp1, tmp2, tmp3, dst0;
__m128i shift1 = {0x0703070307030703, 0x0703070307030703};
__m128i shift2 = {0x0200020002000200, 0x0200020002000200};
for (x = 0; x < len; x++) {
DUP2_ARG2(__lsx_vld, src_argb, 0, src_argb, 16, src0, src1);
tmp0 = __lsx_vpickev_b(src1, src0);
tmp1 = __lsx_vpickod_b(src1, src0);
tmp0 = __lsx_vsrli_b(tmp0, 3);
tmp1 = __lsx_vsrl_b(tmp1, shift1);
tmp0 = __lsx_vsll_b(tmp0, shift2);
tmp2 = __lsx_vpackev_b(zero, tmp1);
tmp3 = __lsx_vpackod_b(zero, tmp1);
tmp2 = __lsx_vslli_h(tmp2, 5);
tmp3 = __lsx_vslli_h(tmp3, 15);
dst0 = __lsx_vor_v(tmp0, tmp2);
dst0 = __lsx_vor_v(dst0, tmp3);
__lsx_vst(dst0, dst_rgb, 0);
dst_rgb += 16;
src_argb += 32;
}
}
void ARGBToARGB4444Row_LSX(const uint8_t* src_argb,
uint8_t* dst_rgb,
int width) {
int x;
int len = width / 8;
__m128i src0, src1, tmp0, tmp1, dst0;
for (x = 0; x < len; x++) {
DUP2_ARG2(__lsx_vld, src_argb, 0, src_argb, 16, src0, src1);
tmp0 = __lsx_vpickev_b(src1, src0);
tmp1 = __lsx_vpickod_b(src1, src0);
tmp1 = __lsx_vandi_b(tmp1, 0xF0);
tmp0 = __lsx_vsrli_b(tmp0, 4);
dst0 = __lsx_vor_v(tmp1, tmp0);
__lsx_vst(dst0, dst_rgb, 0);
dst_rgb += 16;
src_argb += 32;
}
}
void ARGBToUV444Row_LSX(const uint8_t* src_argb,
uint8_t* dst_u,
uint8_t* dst_v,
int32_t width) {
int x;
int len = width / 16;
__m128i src0, src1, src2, src3;
__m128i tmp0, tmp1, tmp2, tmp3;
__m128i reg0, reg1, reg2, reg3, dst0, dst1;
__m128i const_112 = __lsx_vldi(112);
__m128i const_74 = __lsx_vldi(74);
__m128i const_38 = __lsx_vldi(38);
__m128i const_94 = __lsx_vldi(94);
__m128i const_18 = __lsx_vldi(18);
__m128i const_0x8080 = {0x8080808080808080, 0x8080808080808080};
for (x = 0; x < len; x++) {
DUP4_ARG2(__lsx_vld, src_argb, 0, src_argb, 16, src_argb, 32, src_argb,
48, src0, src1, src2, src3);
tmp0 = __lsx_vpickev_h(src1, src0);
tmp1 = __lsx_vpickod_h(src1, src0);
tmp2 = __lsx_vpickev_h(src3, src2);
tmp3 = __lsx_vpickod_h(src3, src2);
reg0 = __lsx_vmaddwev_h_bu(const_0x8080, tmp0, const_112);
reg1 = __lsx_vmaddwev_h_bu(const_0x8080, tmp2, const_112);
reg2 = __lsx_vmulwod_h_bu(tmp0, const_74);
reg3 = __lsx_vmulwod_h_bu(tmp2, const_74);
reg2 = __lsx_vmaddwev_h_bu(reg2, tmp1, const_38);
reg3 = __lsx_vmaddwev_h_bu(reg3, tmp3, const_38);
reg0 = __lsx_vsub_h(reg0, reg2);
reg1 = __lsx_vsub_h(reg1, reg3);
reg0 = __lsx_vsrai_h(reg0, 8);
reg1 = __lsx_vsrai_h(reg1, 8);
dst0 = __lsx_vpickev_b(reg1, reg0);
reg0 = __lsx_vmaddwev_h_bu(const_0x8080, tmp1, const_112);
reg1 = __lsx_vmaddwev_h_bu(const_0x8080, tmp3, const_112);
reg2 = __lsx_vmulwev_h_bu(tmp0, const_18);
reg3 = __lsx_vmulwev_h_bu(tmp2, const_18);
reg2 = __lsx_vmaddwod_h_bu(reg2, tmp0, const_94);
reg3 = __lsx_vmaddwod_h_bu(reg3, tmp2, const_94);
reg0 = __lsx_vsub_h(reg0, reg2);
reg1 = __lsx_vsub_h(reg1, reg3);
reg0 = __lsx_vsrai_h(reg0, 8);
reg1 = __lsx_vsrai_h(reg1, 8);
dst1 = __lsx_vpickev_b(reg1, reg0);
__lsx_vst(dst0, dst_u, 0);
__lsx_vst(dst1, dst_v, 0);
dst_u += 16;
dst_v += 16;
src_argb += 64;
}
}
void ARGBMultiplyRow_LSX(const uint8_t* src_argb0,
const uint8_t* src_argb1,
uint8_t* dst_argb,
int width) {
int x;
int len = width / 4;
__m128i zero = __lsx_vldi(0);
__m128i src0, src1, dst0, dst1;
__m128i tmp0, tmp1, tmp2, tmp3;
for (x = 0; x < len; x++) {
DUP2_ARG2(__lsx_vld, src_argb0, 0, src_argb1, 0, src0, src1);
tmp0 = __lsx_vilvl_b(src0, src0);
tmp1 = __lsx_vilvh_b(src0, src0);
tmp2 = __lsx_vilvl_b(zero, src1);
tmp3 = __lsx_vilvh_b(zero, src1);
dst0 = __lsx_vmuh_hu(tmp0, tmp2);
dst1 = __lsx_vmuh_hu(tmp1, tmp3);
dst0 = __lsx_vpickev_b(dst1, dst0);
__lsx_vst(dst0, dst_argb, 0);
src_argb0 += 16;
src_argb1 += 16;
dst_argb += 16;
}
}
void ARGBAddRow_LSX(const uint8_t* src_argb0,
const uint8_t* src_argb1,
uint8_t* dst_argb,
int width) {
int x;
int len = width / 4;
__m128i src0, src1, dst0;
for (x = 0; x < len; x++) {
DUP2_ARG2(__lsx_vld, src_argb0, 0, src_argb1, 0, src0, src1);
dst0 = __lsx_vsadd_bu(src0, src1);
__lsx_vst(dst0, dst_argb, 0);
src_argb0 += 16;
src_argb1 += 16;
dst_argb += 16;
}
}
void ARGBSubtractRow_LSX(const uint8_t* src_argb0,
const uint8_t* src_argb1,
uint8_t* dst_argb,
int width) {
int x;
int len = width / 4;
__m128i src0, src1, dst0;
for (x = 0; x < len; x++) {
DUP2_ARG2(__lsx_vld, src_argb0, 0, src_argb1, 0, src0, src1);
dst0 = __lsx_vssub_bu(src0, src1);
__lsx_vst(dst0, dst_argb, 0);
src_argb0 += 16;
src_argb1 += 16;
dst_argb += 16;
}
}
void ARGBAttenuateRow_LSX(const uint8_t* src_argb,
uint8_t* dst_argb,
int width) {
int x;
int len = width / 8;
__m128i src0, src1, tmp0, tmp1;
__m128i reg0, reg1, reg2, reg3, reg4, reg5;
__m128i b, g, r, a, dst0, dst1;
__m128i control = {0x0005000100040000, 0x0007000300060002};
for (x = 0; x < len; x++) {
DUP2_ARG2(__lsx_vld, src_argb, 0, src_argb, 16, src0, src1);
tmp0 = __lsx_vpickev_b(src1, src0);
tmp1 = __lsx_vpickod_b(src1, src0);
b = __lsx_vpackev_b(tmp0, tmp0);
r = __lsx_vpackod_b(tmp0, tmp0);
g = __lsx_vpackev_b(tmp1, tmp1);
a = __lsx_vpackod_b(tmp1, tmp1);
reg0 = __lsx_vmulwev_w_hu(b, a);
reg1 = __lsx_vmulwod_w_hu(b, a);
reg2 = __lsx_vmulwev_w_hu(r, a);
reg3 = __lsx_vmulwod_w_hu(r, a);
reg4 = __lsx_vmulwev_w_hu(g, a);
reg5 = __lsx_vmulwod_w_hu(g, a);
reg0 = __lsx_vssrani_h_w(reg1, reg0, 24);
reg2 = __lsx_vssrani_h_w(reg3, reg2, 24);
reg4 = __lsx_vssrani_h_w(reg5, reg4, 24);
reg0 = __lsx_vshuf_h(control, reg0, reg0);
reg2 = __lsx_vshuf_h(control, reg2, reg2);
reg4 = __lsx_vshuf_h(control, reg4, reg4);
tmp0 = __lsx_vpackev_b(reg4, reg0);
tmp1 = __lsx_vpackev_b(a, reg2);
dst0 = __lsx_vilvl_h(tmp1, tmp0);
dst1 = __lsx_vilvh_h(tmp1, tmp0);
__lsx_vst(dst0, dst_argb, 0);
__lsx_vst(dst1, dst_argb, 16);
dst_argb += 32;
src_argb += 32;
}
}
void ARGBToRGB565DitherRow_LSX(const uint8_t* src_argb,
uint8_t* dst_rgb,
const uint32_t dither4,
int width) {
int x;
int len = width / 8;
__m128i src0, src1, tmp0, tmp1, dst0;
__m128i b, g, r;
__m128i zero = __lsx_vldi(0);
__m128i vec_dither = __lsx_vldrepl_w(&dither4, 0);
vec_dither = __lsx_vilvl_b(zero, vec_dither);
for (x = 0; x < len; x++) {
DUP2_ARG2(__lsx_vld, src_argb, 0, src_argb, 16, src0, src1);
tmp0 = __lsx_vpickev_b(src1, src0);
tmp1 = __lsx_vpickod_b(src1, src0);
b = __lsx_vpackev_b(zero, tmp0);
r = __lsx_vpackod_b(zero, tmp0);
g = __lsx_vpackev_b(zero, tmp1);
b = __lsx_vadd_h(b, vec_dither);
g = __lsx_vadd_h(g, vec_dither);
r = __lsx_vadd_h(r, vec_dither);
DUP2_ARG1(__lsx_vclip255_h, b, g, b, g);
r = __lsx_vclip255_h(r);
b = __lsx_vsrai_h(b, 3);
g = __lsx_vsrai_h(g, 2);
r = __lsx_vsrai_h(r, 3);
g = __lsx_vslli_h(g, 5);
r = __lsx_vslli_h(r, 11);
dst0 = __lsx_vor_v(b, g);
dst0 = __lsx_vor_v(dst0, r);
__lsx_vst(dst0, dst_rgb, 0);
src_argb += 32;
dst_rgb += 16;
}
}
void ARGBShuffleRow_LSX(const uint8_t* src_argb,
uint8_t* dst_argb,
const uint8_t* shuffler,
int width) {
int x;
int len = width / 8;
__m128i src0, src1, dst0, dst1;
__m128i shuf = {0x0404040400000000, 0x0C0C0C0C08080808};
__m128i temp = __lsx_vldrepl_w(shuffler, 0);
shuf = __lsx_vadd_b(shuf, temp);
for (x = 0; x < len; x++) {
DUP2_ARG2(__lsx_vld, src_argb, 0, src_argb, 16, src0, src1);
dst0 = __lsx_vshuf_b(src0, src0, shuf);
dst1 = __lsx_vshuf_b(src1, src1, shuf);
__lsx_vst(dst0, dst_argb, 0);
__lsx_vst(dst1, dst_argb, 16);
src_argb += 32;
dst_argb += 32;
}
}
void ARGBShadeRow_LSX(const uint8_t* src_argb,
uint8_t* dst_argb,
int width,
uint32_t value) {
int x;
int len = width / 4;
__m128i src0, dst0, tmp0, tmp1;
__m128i vec_value = __lsx_vreplgr2vr_w(value);
vec_value = __lsx_vilvl_b(vec_value, vec_value);
for (x = 0; x < len; x++) {
src0 = __lsx_vld(src_argb, 0);
tmp0 = __lsx_vilvl_b(src0, src0);
tmp1 = __lsx_vilvh_b(src0, src0);
tmp0 = __lsx_vmuh_hu(tmp0, vec_value);
tmp1 = __lsx_vmuh_hu(tmp1, vec_value);
dst0 = __lsx_vpickod_b(tmp1, tmp0);
__lsx_vst(dst0, dst_argb, 0);
src_argb += 16;
dst_argb += 16;
}
}
void ARGBGrayRow_LSX(const uint8_t* src_argb, uint8_t* dst_argb, int width) {
int x;
int len = width / 8;
__m128i src0, src1, tmp0, tmp1;
__m128i reg0, reg1, reg2, dst0, dst1;
__m128i const_128 = __lsx_vldi(0x480);
__m128i const_150 = __lsx_vldi(0x96);
__m128i const_br = {0x4D1D4D1D4D1D4D1D, 0x4D1D4D1D4D1D4D1D};
for (x = 0; x < len; x++) {
DUP2_ARG2(__lsx_vld, src_argb, 0, src_argb, 16, src0, src1);
tmp0 = __lsx_vpickev_b(src1, src0);
tmp1 = __lsx_vpickod_b(src1, src0);
reg0 = __lsx_vdp2_h_bu(tmp0, const_br);
reg1 = __lsx_vmaddwev_h_bu(const_128, tmp1, const_150);
reg2 = __lsx_vadd_h(reg0, reg1);
tmp0 = __lsx_vpackod_b(reg2, reg2);
tmp1 = __lsx_vpackod_b(tmp1, reg2);
dst0 = __lsx_vilvl_h(tmp1, tmp0);
dst1 = __lsx_vilvh_h(tmp1, tmp0);
__lsx_vst(dst0, dst_argb, 0);
__lsx_vst(dst1, dst_argb, 16);
src_argb += 32;
dst_argb += 32;
}
}
void ARGBSepiaRow_LSX(uint8_t* dst_argb, int width) {
int x;
int len = width / 8;
__m128i src0, src1, tmp0, tmp1;
__m128i reg0, reg1, spb, spg, spr;
__m128i dst0, dst1;
__m128i spb_g = __lsx_vldi(68);
__m128i spg_g = __lsx_vldi(88);
__m128i spr_g = __lsx_vldi(98);
__m128i spb_br = {0x2311231123112311, 0x2311231123112311};
__m128i spg_br = {0x2D162D162D162D16, 0x2D162D162D162D16};
__m128i spr_br = {0x3218321832183218, 0x3218321832183218};
__m128i shuff = {0x1706150413021100, 0x1F0E1D0C1B0A1908};
for (x = 0; x < len; x++) {
DUP2_ARG2(__lsx_vld, dst_argb, 0, dst_argb, 16, src0, src1);
tmp0 = __lsx_vpickev_b(src1, src0);
tmp1 = __lsx_vpickod_b(src1, src0);
DUP2_ARG2(__lsx_vdp2_h_bu, tmp0, spb_br, tmp0, spg_br, spb, spg);
spr = __lsx_vdp2_h_bu(tmp0, spr_br);
spb = __lsx_vmaddwev_h_bu(spb, tmp1, spb_g);
spg = __lsx_vmaddwev_h_bu(spg, tmp1, spg_g);
spr = __lsx_vmaddwev_h_bu(spr, tmp1, spr_g);
spb = __lsx_vsrli_h(spb, 7);
spg = __lsx_vsrli_h(spg, 7);
spr = __lsx_vsrli_h(spr, 7);
spg = __lsx_vsat_hu(spg, 7);
spr = __lsx_vsat_hu(spr, 7);
reg0 = __lsx_vpackev_b(spg, spb);
reg1 = __lsx_vshuf_b(tmp1, spr, shuff);
dst0 = __lsx_vilvl_h(reg1, reg0);
dst1 = __lsx_vilvh_h(reg1, reg0);
__lsx_vst(dst0, dst_argb, 0);
__lsx_vst(dst1, dst_argb, 16);
dst_argb += 32;
}
}
void ARGB4444ToARGBRow_LSX(const uint8_t* src_argb4444,
uint8_t* dst_argb,
int width) {