coffee/libyuv
1
0
Fork 0
mirror of https://chromium.googlesource.com/libyuv/libyuv synced 2025-12-08 01:36:47 +08:00
Code Issues Packages Projects Releases Wiki Activity

Add optimization functions in row_lsx.cc file.

Browse Source 
Optimize 44 functions in source/row_lsx.cc file.
All test cases passed on loongarch platform.

Bug: libyuv:913
Change-Id: Ic80a5751314adc2e9bd435f2bbd928ab017a90f9
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/3351467
Reviewed-by: Frank Barchard <fbarchard@chromium.org>
This commit is contained in:
Hao Chen 2021-12-20 20:14:11 +08:00 committed by Frank Barchard
parent de8ae8c679
commit dfe046d272
12 changed files with 2734 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

329
include/libyuv/row.h
View File

@ -684,6 +684,54 @@ extern "C" {
#define HAS_YUY2TOYROW_MMI
#endif
#if !defined(LIBYUV_DISABLE_LSX) && defined(__loongarch_sx)
#define HAS_ARGB4444TOARGBROW_LSX
#define HAS_ARGB1555TOARGBROW_LSX
#define HAS_RGB565TOARGBROW_LSX
#define HAS_RGB24TOARGBROW_LSX
#define HAS_RAWTOARGBROW_LSX
#define HAS_ARGB1555TOYROW_LSX
#define HAS_ARGB1555TOUVROW_LSX
#define HAS_RGB565TOYROW_LSX
#define HAS_RGB565TOUVROW_LSX
#define HAS_RGB24TOYROW_LSX
#define HAS_RGB24TOUVROW_LSX
#define HAS_RAWTOYROW_LSX
#define HAS_RAWTOUVROW_LSX
#define HAS_NV12TOARGBROW_LSX
#define HAS_NV12TORGB565ROW_LSX
#define HAS_NV21TOARGBROW_LSX
#define HAS_SOBELROW_LSX
#define HAS_SOBELTOPLANEROW_LSX
#define HAS_SOBELXYROW_LSX
#define HAS_ARGBTOYJROW_LSX
#define HAS_BGRATOYROW_LSX
#define HAS_BGRATOUVROW_LSX
#define HAS_ABGRTOYROW_LSX
#define HAS_ABGRTOUVROW_LSX
#define HAS_RGBATOYROW_LSX
#define HAS_RGBATOUVROW_LSX
#define HAS_ARGBTOUVJROW_LSX
#define HAS_I444TOARGBROW_LSX
#define HAS_I400TOARGBROW_LSX
#define HAS_J400TOARGBROW_LSX
#define HAS_YUY2TOARGBROW_LSX
#define HAS_UYVYTOARGBROW_LSX
#define HAS_INTERPOLATEROW_LSX
#define HAS_ARGBSETROW_LSX
#define HAS_RAWTORGB24ROW_LSX
#define HAS_MERGEUVROW_LSX
#define HAS_ARGBEXTRACTALPHAROW_LSX
#define HAS_ARGBBLENDROW_LSX
#define HAS_ARGBQUANTIZEROW_LSX
#define HAS_ARGBCOLORMATRIXROW_LSX
#define HAS_SPLITUVROW_LSX
#define HAS_SETROW_LSX
#define HAS_MIRRORSPLITUVROW_LSX
#define HAS_SOBELXROW_LSX
#define HAS_SOBELYROW_LSX
#define HAS_HALFFLOATROW_LSX
#endif
#if !defined(LIBYUV_DISABLE_LASX) && defined(__loongarch_asx)
#define HAS_I422TOARGBROW_LASX
@ -986,6 +1034,12 @@ void I444ToARGBRow_MMI(const uint8_t* src_y,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width);
void I444ToARGBRow_LSX(const uint8_t* src_y,
const uint8_t* src_u,
const uint8_t* src_v,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width);
void I422ToARGBRow_MSA(const uint8_t* src_y,
const uint8_t* src_u,
@ -1103,6 +1157,30 @@ void UYVYToARGBRow_MSA(const uint8_t* src_uyvy,
const struct YuvConstants* yuvconstants,
int width);
void NV12ToARGBRow_LSX(const uint8_t* src_y,
const uint8_t* src_uv,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width);
void NV12ToRGB565Row_LSX(const uint8_t* src_y,
const uint8_t* src_uv,
uint8_t* dst_rgb565,
const struct YuvConstants* yuvconstants,
int width);
void NV21ToARGBRow_LSX(const uint8_t* src_y,
const uint8_t* src_vu,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width);
void YUY2ToARGBRow_LSX(const uint8_t* src_yuy2,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width);
void UYVYToARGBRow_LSX(const uint8_t* src_uyvy,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width);
void ARGBToYRow_AVX2(const uint8_t* src_argb, uint8_t* dst_y, int width);
void ARGBToYRow_Any_AVX2(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
void ABGRToYRow_AVX2(const uint8_t* src_abgr, uint8_t* dst_y, int width);
@ -1131,6 +1209,7 @@ void ARGBToYJRow_MSA(const uint8_t* src_argb0, uint8_t* dst_y, int width);
void ARGBToYRow_MMI(const uint8_t* src_argb0, uint8_t* dst_y, int width);
void ARGBToYJRow_MMI(const uint8_t* src_argb0, uint8_t* dst_y, int width);
void ARGBToYRow_LASX(const uint8_t* src_argb0, uint8_t* dst_y, int width);
void ARGBToYJRow_LSX(const uint8_t* src_argb0, uint8_t* dst_y, int width);
void ARGBToUV444Row_NEON(const uint8_t* src_argb,
uint8_t* dst_u,
uint8_t* dst_v,
@ -1267,6 +1346,46 @@ void ARGBToUVJRow_MMI(const uint8_t* src_rgb,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void BGRAToUVRow_LSX(const uint8_t* src_bgra,
int src_stride_bgra,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void ABGRToUVRow_LSX(const uint8_t* src_abgr,
int src_stride_abgr,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void RGBAToUVRow_LSX(const uint8_t* src_rgba,
int src_stride_rgba,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void ARGBToUVJRow_LSX(const uint8_t* src_argb,
int src_stride_argb,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void ARGB1555ToUVRow_LSX(const uint8_t* src_argb1555,
int src_stride_argb1555,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void RGB565ToUVRow_LSX(const uint8_t* src_rgb565,
int src_stride_rgb565,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void RGB24ToUVRow_LSX(const uint8_t* src_rgb24,
int src_stride_rgb24,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void RAWToUVRow_LSX(const uint8_t* src_raw,
int src_stride_raw,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void BGRAToUVRow_MMI(const uint8_t* src_rgb,
int src_stride_rgb,
uint8_t* dst_u,
@ -1337,6 +1456,14 @@ void RGB565ToYRow_MMI(const uint8_t* src_rgb565, uint8_t* dst_y, int width);
void ARGB1555ToYRow_MMI(const uint8_t* src_argb1555, uint8_t* dst_y, int width);
void ARGB4444ToYRow_MMI(const uint8_t* src_argb4444, uint8_t* dst_y, int width);
void BGRAToYRow_LSX(const uint8_t* src_bgra, uint8_t* dst_y, int width);
void ABGRToYRow_LSX(const uint8_t* src_abgr, uint8_t* dst_y, int width);
void RGBAToYRow_LSX(const uint8_t* src_rgba, uint8_t* dst_y, int width);
void ARGB1555ToYRow_LSX(const uint8_t* src_argb1555, uint8_t* dst_y, int width);
void RGB565ToYRow_LSX(const uint8_t* src_rgb565, uint8_t* dst_y, int width);
void RGB24ToYRow_LSX(const uint8_t* src_rgb24, uint8_t* dst_y, int width);
void RAWToYRow_LSX(const uint8_t* src_raw, uint8_t* dst_y, int width);
void ARGBToYRow_C(const uint8_t* src_rgb, uint8_t* dst_y, int width);
void ARGBToYJRow_C(const uint8_t* src_rgb, uint8_t* dst_y, int width);
void RGBAToYJRow_C(const uint8_t* src_rgb, uint8_t* dst_y, int width);
@ -1407,7 +1534,17 @@ void ARGB4444ToYRow_Any_MMI(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int width);
void BGRAToYRow_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
void ABGRToYRow_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
void RGBAToYRow_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
void ARGBToYJRow_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
void RGB24ToYRow_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
void RGB565ToYRow_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
void ARGB1555ToYRow_Any_LSX(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int width);
void ARGBToYRow_Any_LASX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
void RAWToYRow_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
void ARGBToUVRow_AVX2(const uint8_t* src_argb,
int src_stride_argb,
@ -1625,6 +1762,46 @@ void ARGBToUVJRow_Any_MMI(const uint8_t* src_ptr,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void ABGRToUVRow_Any_LSX(const uint8_t* src_ptr,
int src_stride_ptr,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void BGRAToUVRow_Any_LSX(const uint8_t* src_ptr,
int src_stride_ptr,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void RGBAToUVRow_Any_LSX(const uint8_t* src_ptr,
int src_stride_ptr,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void ARGBToUVJRow_Any_LSX(const uint8_t* src_ptr,
int src_stride_ptr,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void ARGB1555ToUVRow_Any_LSX(const uint8_t* src_ptr,
int src_stride_ptr,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void RGB565ToUVRow_Any_LSX(const uint8_t* src_ptr,
int src_stride_ptr,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void RGB24ToUVRow_Any_LSX(const uint8_t* src_ptr,
int src_stride_ptr,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void RAWToUVRow_Any_LSX(const uint8_t* src_ptr,
int src_stride_ptr,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void BGRAToUVRow_Any_MMI(const uint8_t* src_ptr,
int src_stride_ptr,
uint8_t* dst_u,
@ -1792,6 +1969,10 @@ void MirrorSplitUVRow_MMI(const uint8_t* src_uv,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void MirrorSplitUVRow_LSX(const uint8_t* src_uv,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void MirrorSplitUVRow_C(const uint8_t* src_uv,
uint8_t* dst_u,
uint8_t* dst_v,
@ -1855,6 +2036,10 @@ void SplitUVRow_MMI(const uint8_t* src_uv,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void SplitUVRow_LSX(const uint8_t* src_uv,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void SplitUVRow_Any_SSE2(const uint8_t* src_ptr,
uint8_t* dst_u,
uint8_t* dst_v,
@ -1875,6 +2060,10 @@ void SplitUVRow_Any_MMI(const uint8_t* src_ptr,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void SplitUVRow_Any_LSX(const uint8_t* src_ptr,
uint8_t* dst_u,
uint8_t* dst_v,
int width);
void MergeUVRow_C(const uint8_t* src_u,
const uint8_t* src_v,
@ -1900,6 +2089,10 @@ void MergeUVRow_MMI(const uint8_t* src_u,
const uint8_t* src_v,
uint8_t* dst_uv,
int width);
void MergeUVRow_LSX(const uint8_t* src_u,
const uint8_t* src_v,
uint8_t* dst_uv,
int width);
void MergeUVRow_Any_SSE2(const uint8_t* y_buf,
const uint8_t* uv_buf,
uint8_t* dst_ptr,
@ -1920,6 +2113,10 @@ void MergeUVRow_Any_MMI(const uint8_t* y_buf,
const uint8_t* uv_buf,
uint8_t* dst_ptr,
int width);
void MergeUVRow_Any_LSX(const uint8_t* y_buf,
const uint8_t* uv_buf,
uint8_t* dst_ptr,
int width);
void HalfMergeUVRow_C(const uint8_t* src_u,
int src_stride_u,
@ -2548,6 +2745,9 @@ void ARGBExtractAlphaRow_MSA(const uint8_t* src_argb,
void ARGBExtractAlphaRow_MMI(const uint8_t* src_argb,
uint8_t* dst_a,
int width);
void ARGBExtractAlphaRow_LSX(const uint8_t* src_argb,
uint8_t* dst_a,
int width);
void ARGBExtractAlphaRow_Any_SSE2(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int width);
@ -2563,6 +2763,9 @@ void ARGBExtractAlphaRow_Any_MSA(const uint8_t* src_ptr,
void ARGBExtractAlphaRow_Any_MMI(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int width);
void ARGBExtractAlphaRow_Any_LSX(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int width);
void ARGBCopyYToAlphaRow_C(const uint8_t* src, uint8_t* dst, int width);
void ARGBCopyYToAlphaRow_SSE2(const uint8_t* src, uint8_t* dst, int width);
@ -2583,8 +2786,10 @@ void SetRow_MSA(uint8_t* dst, uint8_t v8, int width);
void SetRow_X86(uint8_t* dst, uint8_t v8, int width);
void SetRow_ERMS(uint8_t* dst, uint8_t v8, int width);
void SetRow_NEON(uint8_t* dst, uint8_t v8, int width);
void SetRow_LSX(uint8_t* dst, uint8_t v8, int width);
void SetRow_Any_X86(uint8_t* dst_ptr, uint8_t v32, int width);
void SetRow_Any_NEON(uint8_t* dst_ptr, uint8_t v32, int width);
void SetRow_Any_LSX(uint8_t* dst_ptr, uint8_t v32, int width);
void ARGBSetRow_C(uint8_t* dst_argb, uint32_t v32, int width);
void ARGBSetRow_X86(uint8_t* dst_argb, uint32_t v32, int width);
@ -2594,6 +2799,8 @@ void ARGBSetRow_MSA(uint8_t* dst_argb, uint32_t v32, int width);
void ARGBSetRow_Any_MSA(uint8_t* dst_ptr, uint32_t v32, int width);
void ARGBSetRow_MMI(uint8_t* dst_argb, uint32_t v32, int width);
void ARGBSetRow_Any_MMI(uint8_t* dst_ptr, uint32_t v32, int width);
void ARGBSetRow_LSX(uint8_t* dst_argb, uint32_t v32, int width);
void ARGBSetRow_Any_LSX(uint8_t* dst_ptr, uint32_t v32, int width);
// ARGBShufflers for BGRAToARGB etc.
void ARGBShuffleRow_C(const uint8_t* src_argb,
@ -2673,13 +2880,16 @@ void RGB24ToARGBRow_NEON(const uint8_t* src_rgb24,
int width);
void RGB24ToARGBRow_MSA(const uint8_t* src_rgb24, uint8_t* dst_argb, int width);
void RGB24ToARGBRow_MMI(const uint8_t* src_rgb24, uint8_t* dst_argb, int width);
void RGB24ToARGBRow_LSX(const uint8_t* src_rgb24, uint8_t* dst_argb, int width);
void RAWToARGBRow_NEON(const uint8_t* src_raw, uint8_t* dst_argb, int width);
void RAWToRGBARow_NEON(const uint8_t* src_raw, uint8_t* dst_rgba, int width);
void RAWToARGBRow_MSA(const uint8_t* src_raw, uint8_t* dst_argb, int width);
void RAWToARGBRow_MMI(const uint8_t* src_raw, uint8_t* dst_argb, int width);
void RAWToARGBRow_LSX(const uint8_t* src_raw, uint8_t* dst_argb, int width);
void RAWToRGB24Row_NEON(const uint8_t* src_raw, uint8_t* dst_rgb24, int width);
void RAWToRGB24Row_MSA(const uint8_t* src_raw, uint8_t* dst_rgb24, int width);
void RAWToRGB24Row_MMI(const uint8_t* src_raw, uint8_t* dst_rgb24, int width);
void RAWToRGB24Row_LSX(const uint8_t* src_raw, uint8_t* dst_rgb24, int width);
void RGB565ToARGBRow_NEON(const uint8_t* src_rgb565,
uint8_t* dst_argb,
int width);
@ -2689,6 +2899,9 @@ void RGB565ToARGBRow_MSA(const uint8_t* src_rgb565,
void RGB565ToARGBRow_MMI(const uint8_t* src_rgb565,
uint8_t* dst_argb,
int width);
void RGB565ToARGBRow_LSX(const uint8_t* src_rgb565,
uint8_t* dst_argb,
int width);
void ARGB1555ToARGBRow_NEON(const uint8_t* src_argb1555,
uint8_t* dst_argb,
int width);
@ -2698,6 +2911,9 @@ void ARGB1555ToARGBRow_MSA(const uint8_t* src_argb1555,
void ARGB1555ToARGBRow_MMI(const uint8_t* src_argb1555,
uint8_t* dst_argb,
int width);
void ARGB1555ToARGBRow_LSX(const uint8_t* src_argb1555,
uint8_t* dst_argb,
int width);
void ARGB4444ToARGBRow_NEON(const uint8_t* src_argb4444,
uint8_t* dst_argb,
int width);
@ -2707,6 +2923,9 @@ void ARGB4444ToARGBRow_MSA(const uint8_t* src_argb4444,
void ARGB4444ToARGBRow_MMI(const uint8_t* src_argb4444,
uint8_t* dst_argb,
int width);
void ARGB4444ToARGBRow_LSX(const uint8_t* src_argb4444,
uint8_t* dst_argb,
int width);
void RGB24ToARGBRow_C(const uint8_t* src_rgb24, uint8_t* dst_argb, int width);
void RAWToARGBRow_C(const uint8_t* src_raw, uint8_t* dst_argb, int width);
void RAWToRGBARow_C(const uint8_t* src_raw, uint8_t* dst_rgba, int width);
@ -2764,15 +2983,20 @@ void RGB24ToARGBRow_Any_MSA(const uint8_t* src_ptr,
void RGB24ToARGBRow_Any_MMI(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int width);
void RGB24ToARGBRow_Any_LSX(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int width);
void RAWToARGBRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
void RAWToRGBARow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
void RAWToARGBRow_Any_MSA(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
void RAWToARGBRow_Any_MMI(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
void RAWToARGBRow_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
void RAWToRGB24Row_Any_NEON(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int width);
void RAWToRGB24Row_Any_MSA(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
void RAWToRGB24Row_Any_MMI(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
void RAWToRGB24Row_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
void RGB565ToARGBRow_Any_NEON(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int width);
@ -2782,6 +3006,9 @@ void RGB565ToARGBRow_Any_MSA(const uint8_t* src_ptr,
void RGB565ToARGBRow_Any_MMI(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int width);
void RGB565ToARGBRow_Any_LSX(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int width);
void ARGB1555ToARGBRow_Any_NEON(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int width);
@ -2794,6 +3021,9 @@ void ARGB1555ToARGBRow_Any_MMI(const uint8_t* src_ptr,
void ARGB4444ToARGBRow_Any_NEON(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int width);
void ARGB1555ToARGBRow_Any_LSX(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int width);
void ARGB4444ToARGBRow_Any_MSA(const uint8_t* src_ptr,
uint8_t* dst_ptr,
@ -2801,6 +3031,9 @@ void ARGB4444ToARGBRow_Any_MSA(const uint8_t* src_ptr,
void ARGB4444ToARGBRow_Any_MMI(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int width);
void ARGB4444ToARGBRow_Any_LSX(const uint8_t* src_ptr,
uint8_t* dst_ptr,
int width);
void ARGBToRGB24Row_SSSE3(const uint8_t* src, uint8_t* dst, int width);
void ARGBToRAWRow_SSSE3(const uint8_t* src, uint8_t* dst, int width);
@ -2976,6 +3209,7 @@ void J400ToARGBRow_AVX2(const uint8_t* src_y, uint8_t* dst_argb, int width);
void J400ToARGBRow_NEON(const uint8_t* src_y, uint8_t* dst_argb, int width);
void J400ToARGBRow_MSA(const uint8_t* src_y, uint8_t* dst_argb, int width);
void J400ToARGBRow_MMI(const uint8_t* src_y, uint8_t* dst_argb, int width);
void J400ToARGBRow_LSX(const uint8_t* src_y, uint8_t* dst_argb, int width);
void J400ToARGBRow_C(const uint8_t* src_y, uint8_t* dst_argb, int width);
void J400ToARGBRow_Any_SSE2(const uint8_t* src_ptr,
uint8_t* dst_ptr,
@ -2988,6 +3222,7 @@ void J400ToARGBRow_Any_NEON(const uint8_t* src_ptr,
int width);
void J400ToARGBRow_Any_MSA(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
void J400ToARGBRow_Any_MMI(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
void J400ToARGBRow_Any_LSX(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
void I444ToARGBRow_C(const uint8_t* src_y,
const uint8_t* src_u,
@ -3875,6 +4110,10 @@ void I400ToARGBRow_MMI(const uint8_t* src_y,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width);
void I400ToARGBRow_LSX(const uint8_t* src_y,
uint8_t* dst_argb,
const struct YuvConstants* yuvconstants,
int width);
void I400ToARGBRow_Any_SSE2(const uint8_t* src_ptr,
uint8_t* dst_ptr,
const struct YuvConstants* param,
@ -3895,6 +4134,10 @@ void I400ToARGBRow_Any_MMI(const uint8_t* src_ptr,
uint8_t* dst_ptr,
const struct YuvConstants* yuvconstants,
int width);
void I400ToARGBRow_Any_LSX(const uint8_t* src_ptr,
uint8_t* dst_ptr,
const struct YuvConstants* yuvconstants,
int width);
// ARGB preattenuated alpha blend.
void ARGBBlendRow_SSSE3(const uint8_t* src_argb,
@ -3913,6 +4156,10 @@ void ARGBBlendRow_MMI(const uint8_t* src_argb0,
const uint8_t* src_argb1,
uint8_t* dst_argb,
int width);
void ARGBBlendRow_LSX(const uint8_t* src_argb0,
const uint8_t* src_argb1,
uint8_t* dst_argb,
int width);
void ARGBBlendRow_C(const uint8_t* src_argb,
const uint8_t* src_argb1,
uint8_t* dst_argb,
@ -4388,6 +4635,12 @@ void I444ToARGBRow_Any_MMI(const uint8_t* y_buf,
uint8_t* dst_ptr,
const struct YuvConstants* yuvconstants,
int width);
void I444ToARGBRow_Any_LSX(const uint8_t* y_buf,
const uint8_t* u_buf,
const uint8_t* v_buf,
uint8_t* dst_ptr,
const struct YuvConstants* yuvconstants,
int width);
void I422ToARGBRow_Any_MSA(const uint8_t* y_buf,
const uint8_t* u_buf,
const uint8_t* v_buf,
@ -4504,6 +4757,30 @@ void UYVYToARGBRow_Any_MSA(const uint8_t* src_ptr,
const struct YuvConstants* yuvconstants,
int width);
void NV12ToARGBRow_Any_LSX(const uint8_t* y_buf,
const uint8_t* uv_buf,
uint8_t* dst_ptr,
const struct YuvConstants* yuvconstants,
int width);
void NV12ToRGB565Row_Any_LSX(const uint8_t* y_buf,
const uint8_t* uv_buf,
uint8_t* dst_ptr,
const struct YuvConstants* yuvconstants,
int width);
void NV21ToARGBRow_Any_LSX(const uint8_t* y_buf,
const uint8_t* uv_buf,
uint8_t* dst_ptr,
const struct YuvConstants* yuvconstants,
int width);
void YUY2ToARGBRow_Any_LSX(const uint8_t* src_ptr,
uint8_t* dst_ptr,
const struct YuvConstants* yuvconstants,
int width);
void UYVYToARGBRow_Any_LSX(const uint8_t* src_ptr,
uint8_t* dst_ptr,
const struct YuvConstants* yuvconstants,
int width);
void YUY2ToYRow_AVX2(const uint8_t* src_yuy2, uint8_t* dst_y, int width);
void YUY2ToUVRow_AVX2(const uint8_t* src_yuy2,
int stride_yuy2,
@ -5032,6 +5309,10 @@ void ARGBColorMatrixRow_MMI(const uint8_t* src_argb,
uint8_t* dst_argb,
const int8_t* matrix_argb,
int width);
void ARGBColorMatrixRow_LSX(const uint8_t* src_argb,
uint8_t* dst_argb,
const int8_t* matrix_argb,
int width);
void ARGBColorTableRow_C(uint8_t* dst_argb,
const uint8_t* table_argb,
@ -5067,6 +5348,12 @@ void ARGBQuantizeRow_MSA(uint8_t* dst_argb,
int interval_size,
int interval_offset,
int width);
void ARGBQuantizeRow_LSX(uint8_t* dst_argb,
int scale,
int interval_size,
int interval_offset,
int width);
void ARGBShadeRow_C(const uint8_t* src_argb,
uint8_t* dst_argb,
@ -5165,6 +5452,11 @@ void InterpolateRow_MMI(uint8_t* dst_ptr,
ptrdiff_t src_stride,
int width,
int source_y_fraction);
void InterpolateRow_LSX(uint8_t* dst_ptr,
const uint8_t* src_ptr,
ptrdiff_t src_stride,
int width,
int source_y_fraction);
void InterpolateRow_Any_NEON(uint8_t* dst_ptr,
const uint8_t* src_ptr,
ptrdiff_t src_stride_ptr,
@ -5190,6 +5482,11 @@ void InterpolateRow_Any_MMI(uint8_t* dst_ptr,
ptrdiff_t src_stride_ptr,
int width,
int source_y_fraction);
void InterpolateRow_Any_LSX(uint8_t* dst_ptr,
const uint8_t* src_ptr,
ptrdiff_t src_stride_ptr,
int width,
int source_y_fraction);
void InterpolateRow_16_C(uint16_t* dst_ptr,
const uint16_t* src_ptr,
@ -5263,6 +5560,10 @@ void SobelRow_MMI(const uint8_t* src_sobelx,
const uint8_t* src_sobely,
uint8_t* dst_argb,
int width);
void SobelRow_LSX(const uint8_t* src_sobelx,
const uint8_t* src_sobely,
uint8_t* dst_argb,
int width);
void SobelToPlaneRow_C(const uint8_t* src_sobelx,
const uint8_t* src_sobely,
uint8_t* dst_y,
@ -5283,6 +5584,10 @@ void SobelToPlaneRow_MMI(const uint8_t* src_sobelx,
const uint8_t* src_sobely,
uint8_t* dst_y,
int width);
void SobelToPlaneRow_LSX(const uint8_t* src_sobelx,
const uint8_t* src_sobely,
uint8_t* dst_y,
int width);
void SobelXYRow_C(const uint8_t* src_sobelx,
const uint8_t* src_sobely,
uint8_t* dst_argb,
@ -5303,6 +5608,10 @@ void SobelXYRow_MMI(const uint8_t* src_sobelx,
const uint8_t* src_sobely,
uint8_t* dst_argb,
int width);
void SobelXYRow_LSX(const uint8_t* src_sobelx,
const uint8_t* src_sobely,
uint8_t* dst_argb,
int width);
void SobelRow_Any_SSE2(const uint8_t* y_buf,
const uint8_t* uv_buf,
uint8_t* dst_ptr,
@ -5319,6 +5628,10 @@ void SobelRow_Any_MMI(const uint8_t* y_buf,
const uint8_t* uv_buf,
uint8_t* dst_ptr,
int width);
void SobelRow_Any_LSX(const uint8_t* y_buf,
const uint8_t* uv_buf,
uint8_t* dst_ptr,
int width);
void SobelToPlaneRow_Any_SSE2(const uint8_t* y_buf,
const uint8_t* uv_buf,
uint8_t* dst_ptr,
@ -5335,6 +5648,10 @@ void SobelToPlaneRow_Any_MMI(const uint8_t* y_buf,
const uint8_t* uv_buf,
uint8_t* dst_ptr,
int width);
void SobelToPlaneRow_Any_LSX(const uint8_t* y_buf,
const uint8_t* uv_buf,
uint8_t* dst_ptr,
int width);
void SobelXYRow_Any_SSE2(const uint8_t* y_buf,
const uint8_t* uv_buf,
uint8_t* dst_ptr,
@ -5351,6 +5668,10 @@ void SobelXYRow_Any_MMI(const uint8_t* y_buf,
const uint8_t* uv_buf,
uint8_t* dst_ptr,
int width);
void SobelXYRow_Any_LSX(const uint8_t* y_buf,
const uint8_t* uv_buf,
uint8_t* dst_ptr,
int width);
void ARGBPolynomialRow_C(const uint8_t* src_argb,
uint8_t* dst_argb,
@ -5423,6 +5744,14 @@ void HalfFloatRow_Any_MSA(const uint16_t* src_ptr,
uint16_t* dst_ptr,
float param,
int width);
void HalfFloatRow_LSX(const uint16_t* src,
uint16_t* dst,
float scale,
int width);
void HalfFloatRow_Any_LSX(const uint16_t* src_ptr,
uint16_t* dst_ptr,
float param,
int width);
void ByteToFloatRow_C(const uint8_t* src, float* dst, float scale, int width);
void ByteToFloatRow_NEON(const uint8_t* src,
float* dst,

120
source/convert.cc
View File

@ -644,6 +644,14 @@ int I422ToNV21(const uint8_t* src_y,
}
}
#endif
#if defined(HAS_MERGEUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
MergeUVRow = MergeUVRow_Any_LSX;
if (IS_ALIGNED(halfwidth, 16)) {
MergeUVRow = MergeUVRow_LSX;
}
}
#endif
#if defined(HAS_INTERPOLATEROW_SSSE3)
if (TestCpuFlag(kCpuHasSSSE3)) {
InterpolateRow = InterpolateRow_Any_SSSE3;
@ -684,6 +692,14 @@ int I422ToNV21(const uint8_t* src_y,
}
}
#endif
#if defined(HAS_INTERPOLATEROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
InterpolateRow = InterpolateRow_Any_LSX;
if (IS_ALIGNED(width, 32)) {
InterpolateRow = InterpolateRow_LSX;
}
}
#endif
if (dst_y) {
CopyPlane(src_y, src_stride_y, dst_y, dst_stride_y, halfwidth, height);
@ -1562,6 +1578,16 @@ int BGRAToI420(const uint8_t* src_bgra,
}
}
#endif
#if defined(HAS_BGRATOYROW_LSX) && defined(HAS_BGRATOUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
BGRAToYRow = BGRAToYRow_Any_LSX;
BGRAToUVRow = BGRAToUVRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
BGRAToYRow = BGRAToYRow_LSX;
BGRAToUVRow = BGRAToUVRow_LSX;
}
}
#endif
for (y = 0; y < height - 1; y += 2) {
BGRAToUVRow(src_bgra, src_stride_bgra, dst_u, dst_v, width);
@ -1676,6 +1702,16 @@ int ABGRToI420(const uint8_t* src_abgr,
}
}
#endif
#if defined(HAS_ABGRTOYROW_LSX) && defined(HAS_ABGRTOUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ABGRToYRow = ABGRToYRow_Any_LSX;
ABGRToUVRow = ABGRToUVRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
ABGRToYRow = ABGRToYRow_LSX;
ABGRToUVRow = ABGRToUVRow_LSX;
}
}
#endif
for (y = 0; y < height - 1; y += 2) {
ABGRToUVRow(src_abgr, src_stride_abgr, dst_u, dst_v, width);
@ -1774,6 +1810,16 @@ int RGBAToI420(const uint8_t* src_rgba,
}
}
#endif
#if defined(HAS_RGBATOYROW_LSX) && defined(HAS_RGBATOUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
RGBAToYRow = RGBAToYRow_Any_LSX;
RGBAToUVRow = RGBAToUVRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
RGBAToYRow = RGBAToYRow_LSX;
RGBAToUVRow = RGBAToUVRow_LSX;
}
}
#endif
for (y = 0; y < height - 1; y += 2) {
RGBAToUVRow(src_rgba, src_stride_rgba, dst_u, dst_v, width);
@ -1793,7 +1839,7 @@ int RGBAToI420(const uint8_t* src_rgba,
// Enabled if 1 pass is available
#if (defined(HAS_RGB24TOYROW_NEON) || defined(HAS_RGB24TOYROW_MSA) || \
defined(HAS_RGB24TOYROW_MMI))
defined(HAS_RGB24TOYROW_MMI) || defined(HAS_RGB24TOYROW_LSX))
#define HAS_RGB24TOYROW
#endif
@ -1872,6 +1918,16 @@ int RGB24ToI420(const uint8_t* src_rgb24,
}
}
#endif
#if defined(HAS_RGB24TOYROW_LSX) && defined(HAS_RGB24TOUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
RGB24ToUVRow = RGB24ToUVRow_Any_LSX;
RGB24ToYRow = RGB24ToYRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
RGB24ToYRow = RGB24ToYRow_LSX;
RGB24ToUVRow = RGB24ToUVRow_LSX;
}
}
#endif
// Other platforms do intermediate conversion from RGB24 to ARGB.
#else // HAS_RGB24TOYROW
@ -2131,7 +2187,7 @@ int RGB24ToJ420(const uint8_t* src_rgb24,
// Enabled if 1 pass is available
#if (defined(HAS_RAWTOYROW_NEON) || defined(HAS_RAWTOYROW_MSA) || \
defined(HAS_RAWTOYROW_MMI))
defined(HAS_RAWTOYROW_MMI) || defined(HAS_RAWTOYROW_LSX))
#define HAS_RAWTOYROW
#endif
@ -2209,6 +2265,16 @@ int RAWToI420(const uint8_t* src_raw,
}
}
#endif
#if defined(HAS_RAWTOYROW_LSX) && defined(HAS_RAWTOUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
RAWToUVRow = RAWToUVRow_Any_LSX;
RAWToYRow = RAWToYRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
RAWToYRow = RAWToYRow_LSX;
RAWToUVRow = RAWToUVRow_LSX;
}
}
#endif
// Other platforms do intermediate conversion from RAW to ARGB.
#else // HAS_RAWTOYROW
@ -2480,7 +2546,7 @@ int RGB565ToI420(const uint8_t* src_rgb565,
int height) {
int y;
#if (defined(HAS_RGB565TOYROW_NEON) || defined(HAS_RGB565TOYROW_MSA) || \
defined(HAS_RGB565TOYROW_MMI))
defined(HAS_RGB565TOYROW_MMI) || defined(HAS_RGB565TOYROW_LSX))
void (*RGB565ToUVRow)(const uint8_t* src_rgb565, int src_stride_rgb565,
uint8_t* dst_u, uint8_t* dst_v, int width) =
RGB565ToUVRow_C;
@ -2518,7 +2584,8 @@ int RGB565ToI420(const uint8_t* src_rgb565,
}
}
// MMI and MSA version does direct RGB565 to YUV.
#elif (defined(HAS_RGB565TOYROW_MMI) || defined(HAS_RGB565TOYROW_MSA))
#elif (defined(HAS_RGB565TOYROW_MMI) || defined(HAS_RGB565TOYROW_MSA) \
|| defined(HAS_RGB565TOYROW_LSX))
#if defined(HAS_RGB565TOYROW_MMI) && defined(HAS_RGB565TOUVROW_MMI)
if (TestCpuFlag(kCpuHasMMI)) {
RGB565ToUVRow = RGB565ToUVRow_Any_MMI;
@ -2541,6 +2608,16 @@ int RGB565ToI420(const uint8_t* src_rgb565,
}
}
#endif
#if defined(HAS_RGB565TOYROW_LSX) && defined(HAS_RGB565TOUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
RGB565ToUVRow = RGB565ToUVRow_Any_LSX;
RGB565ToYRow = RGB565ToYRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
RGB565ToYRow = RGB565ToYRow_LSX;
RGB565ToUVRow = RGB565ToUVRow_LSX;
}
}
#endif
// Other platforms do intermediate conversion from RGB565 to ARGB.
#else
#if defined(HAS_RGB565TOARGBROW_SSE2)
@ -2594,14 +2671,14 @@ int RGB565ToI420(const uint8_t* src_rgb565,
#endif
{
#if !(defined(HAS_RGB565TOYROW_NEON) || defined(HAS_RGB565TOYROW_MSA) || \
defined(HAS_RGB565TOYROW_MMI))
defined(HAS_RGB565TOYROW_MMI) || defined(HAS_RGB565TOYROW_LSX))
// Allocate 2 rows of ARGB.
const int kRowSize = (width * 4 + 31) & ~31;
align_buffer_64(row, kRowSize * 2);
#endif
for (y = 0; y < height - 1; y += 2) {
#if (defined(HAS_RGB565TOYROW_NEON) || defined(HAS_RGB565TOYROW_MSA) || \
defined(HAS_RGB565TOYROW_MMI))
defined(HAS_RGB565TOYROW_MMI) || defined(HAS_RGB565TOYROW_LSX))
RGB565ToUVRow(src_rgb565, src_stride_rgb565, dst_u, dst_v, width);
RGB565ToYRow(src_rgb565, dst_y, width);
RGB565ToYRow(src_rgb565 + src_stride_rgb565, dst_y + dst_stride_y, width);
@ -2619,7 +2696,7 @@ int RGB565ToI420(const uint8_t* src_rgb565,
}
if (height & 1) {
#if (defined(HAS_RGB565TOYROW_NEON) || defined(HAS_RGB565TOYROW_MSA) || \
defined(HAS_RGB565TOYROW_MMI))
defined(HAS_RGB565TOYROW_MMI) || defined(HAS_RGB565TOYROW_LSX))
RGB565ToUVRow(src_rgb565, 0, dst_u, dst_v, width);
RGB565ToYRow(src_rgb565, dst_y, width);
#else
@ -2629,7 +2706,7 @@ int RGB565ToI420(const uint8_t* src_rgb565,
#endif
}
#if !(defined(HAS_RGB565TOYROW_NEON) || defined(HAS_RGB565TOYROW_MSA) || \
defined(HAS_RGB565TOYROW_MMI))
defined(HAS_RGB565TOYROW_MMI) || defined(HAS_RGB565TOYROW_LSX))
free_aligned_buffer_64(row);
#endif
}
@ -2650,7 +2727,7 @@ int ARGB1555ToI420(const uint8_t* src_argb1555,
int height) {
int y;
#if (defined(HAS_ARGB1555TOYROW_NEON) || defined(HAS_ARGB1555TOYROW_MSA) || \
defined(HAS_ARGB1555TOYROW_MMI))
defined(HAS_ARGB1555TOYROW_MMI) || defined(HAS_ARGB1555TOYROW_LSX))
void (*ARGB1555ToUVRow)(const uint8_t* src_argb1555, int src_stride_argb1555,
uint8_t* dst_u, uint8_t* dst_v, int width) =
ARGB1555ToUVRow_C;
@ -2712,6 +2789,15 @@ int ARGB1555ToI420(const uint8_t* src_argb1555,
}
}
#endif
#elif (defined(HAS_ARGB1555TOYROW_LSX) && defined(HAS_ARGB1555TOUVROW_LSX))
if (TestCpuFlag(kCpuHasLSX)) {
ARGB1555ToUVRow = ARGB1555ToUVRow_Any_LSX;
ARGB1555ToYRow = ARGB1555ToYRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
ARGB1555ToYRow = ARGB1555ToYRow_LSX;
ARGB1555ToUVRow = ARGB1555ToUVRow_LSX;
}
}
// Other platforms do intermediate conversion from ARGB1555 to ARGB.
#else
#if defined(HAS_ARGB1555TOARGBROW_SSE2)
@ -2765,7 +2851,7 @@ int ARGB1555ToI420(const uint8_t* src_argb1555,
#endif
{
#if !(defined(HAS_ARGB1555TOYROW_NEON) || defined(HAS_ARGB1555TOYROW_MSA) || \
defined(HAS_ARGB1555TOYROW_MMI))
defined(HAS_ARGB1555TOYROW_MMI) || defined(HAS_ARGB1555TOYROW_LSX))
// Allocate 2 rows of ARGB.
const int kRowSize = (width * 4 + 31) & ~31;
align_buffer_64(row, kRowSize * 2);
@ -2773,7 +2859,7 @@ int ARGB1555ToI420(const uint8_t* src_argb1555,
for (y = 0; y < height - 1; y += 2) {
#if (defined(HAS_ARGB1555TOYROW_NEON) || defined(HAS_ARGB1555TOYROW_MSA) || \
defined(HAS_ARGB1555TOYROW_MMI))
defined(HAS_ARGB1555TOYROW_MMI) || defined(HAS_ARGB1555TOYROW_LSX))
ARGB1555ToUVRow(src_argb1555, src_stride_argb1555, dst_u, dst_v, width);
ARGB1555ToYRow(src_argb1555, dst_y, width);
ARGB1555ToYRow(src_argb1555 + src_stride_argb1555, dst_y + dst_stride_y,
@ -2793,7 +2879,7 @@ int ARGB1555ToI420(const uint8_t* src_argb1555,
}
if (height & 1) {
#if (defined(HAS_ARGB1555TOYROW_NEON) || defined(HAS_ARGB1555TOYROW_MSA) || \
defined(HAS_ARGB1555TOYROW_MMI))
defined(HAS_ARGB1555TOYROW_MMI) || defined(HAS_ARGB1555TOYROW_LSX))
ARGB1555ToUVRow(src_argb1555, 0, dst_u, dst_v, width);
ARGB1555ToYRow(src_argb1555, dst_y, width);
#else
@ -2803,7 +2889,7 @@ int ARGB1555ToI420(const uint8_t* src_argb1555,
#endif
}
#if !(defined(HAS_ARGB1555TOYROW_NEON) || defined(HAS_ARGB1555TOYROW_MSA) || \
defined(HAS_ARGB1555TOYROW_MMI))
defined(HAS_ARGB1555TOYROW_MMI) || defined(HAS_ARGB1555TOYROW_LSX))
free_aligned_buffer_64(row);
#endif
}
@ -2898,6 +2984,14 @@ int ARGB4444ToI420(const uint8_t* src_argb4444,
}
}
#endif
#if defined(HAS_ARGB4444TOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGB4444ToARGBRow = ARGB4444ToARGBRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
ARGB4444ToARGBRow = ARGB4444ToARGBRow_LSX;
}
}
#endif
#if defined(HAS_ARGBTOYROW_SSSE3)
if (TestCpuFlag(kCpuHasSSSE3)) {
ARGBToYRow = ARGBToYRow_Any_SSSE3;

104
source/convert_argb.cc
View File

@ -605,6 +605,14 @@ int I444ToARGBMatrix(const uint8_t* src_y,
}
}
#endif
#if defined(HAS_I444TOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
I444ToARGBRow = I444ToARGBRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
I444ToARGBRow = I444ToARGBRow_LSX;
}
}
#endif
for (y = 0; y < height; ++y) {
I444ToARGBRow(src_y, src_u, src_v, dst_argb, yuvconstants, width);
@ -2731,6 +2739,14 @@ int I400ToARGBMatrix(const uint8_t* src_y,
}
}
#endif
#if defined(HAS_I400TOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
I400ToARGBRow = I400ToARGBRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
I400ToARGBRow = I400ToARGBRow_LSX;
}
}
#endif
for (y = 0; y < height; ++y) {
I400ToARGBRow(src_y, dst_argb, yuvconstants, width);
@ -2817,6 +2833,14 @@ int J400ToARGB(const uint8_t* src_y,
J400ToARGBRow = J400ToARGBRow_MSA;
}
}
#endif
#if defined(HAS_J400TOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
J400ToARGBRow = J400ToARGBRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
J400ToARGBRow = J400ToARGBRow_LSX;
}
}
#endif
for (y = 0; y < height; ++y) {
J400ToARGBRow(src_y, dst_argb, width);
@ -2972,6 +2996,14 @@ int RGB24ToARGB(const uint8_t* src_rgb24,
}
}
#endif
#if defined(HAS_RGB24TOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
RGB24ToARGBRow = RGB24ToARGBRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
RGB24ToARGBRow = RGB24ToARGBRow_LSX;
}
}
#endif
for (y = 0; y < height; ++y) {
RGB24ToARGBRow(src_rgb24, dst_argb, width);
@ -3039,6 +3071,14 @@ int RAWToARGB(const uint8_t* src_raw,
}
}
#endif
#if defined(HAS_RAWTOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
RAWToARGBRow = RAWToARGBRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
RAWToARGBRow = RAWToARGBRow_LSX;
}
}
#endif
for (y = 0; y < height; ++y) {
RAWToARGBRow(src_raw, dst_argb, width);
@ -3165,6 +3205,14 @@ int RGB565ToARGB(const uint8_t* src_rgb565,
}
}
#endif
#if defined(HAS_RGB565TOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
RGB565ToARGBRow = RGB565ToARGBRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
RGB565ToARGBRow = RGB565ToARGBRow_LSX;
}
}
#endif
for (y = 0; y < height; ++y) {
RGB565ToARGBRow(src_rgb565, dst_argb, width);
@ -3240,6 +3288,14 @@ int ARGB1555ToARGB(const uint8_t* src_argb1555,
}
}
#endif
#if defined(HAS_ARGB1555TOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGB1555ToARGBRow = ARGB1555ToARGBRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
ARGB1555ToARGBRow = ARGB1555ToARGBRow_LSX;
}
}
#endif
for (y = 0; y < height; ++y) {
ARGB1555ToARGBRow(src_argb1555, dst_argb, width);
@ -3315,6 +3371,14 @@ int ARGB4444ToARGB(const uint8_t* src_argb4444,
}
}
#endif
#if defined(HAS_ARGB4444TOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGB4444ToARGBRow = ARGB4444ToARGBRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
ARGB4444ToARGBRow = ARGB4444ToARGBRow_LSX;
}
}
#endif
for (y = 0; y < height; ++y) {
ARGB4444ToARGBRow(src_argb4444, dst_argb, width);
@ -3602,6 +3666,14 @@ int NV12ToARGBMatrix(const uint8_t* src_y,
}
}
#endif
#if defined(HAS_NV12TOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
NV12ToARGBRow = NV12ToARGBRow_Any_LSX;
if (IS_ALIGNED(width, 8)) {
NV12ToARGBRow = NV12ToARGBRow_LSX;
}
}
#endif
for (y = 0; y < height; ++y) {
NV12ToARGBRow(src_y, src_uv, dst_argb, yuvconstants, width);
@ -3678,6 +3750,14 @@ int NV21ToARGBMatrix(const uint8_t* src_y,
}
}
#endif
#if defined(HAS_NV21TOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
NV21ToARGBRow = NV21ToARGBRow_Any_LSX;
if (IS_ALIGNED(width, 8)) {
NV21ToARGBRow = NV21ToARGBRow_LSX;
}
}
#endif
for (y = 0; y < height; ++y) {
NV21ToARGBRow(src_y, src_vu, dst_argb, yuvconstants, width);
@ -4065,6 +4145,14 @@ int YUY2ToARGB(const uint8_t* src_yuy2,
YUY2ToARGBRow = YUY2ToARGBRow_MSA;
}
}
#endif
#if defined(HAS_YUY2TOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
YUY2ToARGBRow = YUY2ToARGBRow_Any_LSX;
if (IS_ALIGNED(width, 8)) {
YUY2ToARGBRow = YUY2ToARGBRow_LSX;
}
}
#endif
for (y = 0; y < height; ++y) {
YUY2ToARGBRow(src_yuy2, dst_argb, &kYuvI601Constants, width);
@ -4140,6 +4228,14 @@ int UYVYToARGB(const uint8_t* src_uyvy,
UYVYToARGBRow = UYVYToARGBRow_MSA;
}
}
#endif
#if defined(HAS_UYVYTOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
UYVYToARGBRow = UYVYToARGBRow_Any_LSX;
if (IS_ALIGNED(width, 8)) {
UYVYToARGBRow = UYVYToARGBRow_LSX;
}
}
#endif
for (y = 0; y < height; ++y) {
UYVYToARGBRow(src_uyvy, dst_argb, &kYuvI601Constants, width);
@ -4450,6 +4546,14 @@ int NV12ToRGB565Matrix(const uint8_t* src_y,
}
}
#endif
#if defined(HAS_NV12TORGB565ROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
NV12ToRGB565Row = NV12ToRGB565Row_Any_LSX;
if (IS_ALIGNED(width, 8)) {
NV12ToRGB565Row = NV12ToRGB565Row_LSX;
}
}
#endif
for (y = 0; y < height; ++y) {
NV12ToRGB565Row(src_y, src_uv, dst_rgb565, yuvconstants, width);

52
source/convert_from_argb.cc
View File

@ -431,6 +431,14 @@ int ARGBToNV12(const uint8_t* src_argb,
MergeUVRow_ = MergeUVRow_MSA;
}
}
#endif
#if defined(HAS_MERGEUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
MergeUVRow_ = MergeUVRow_Any_LSX;
if (IS_ALIGNED(halfwidth, 16)) {
MergeUVRow_ = MergeUVRow_LSX;
}
}
#endif
{
// Allocate a rows of uv.
@ -605,6 +613,14 @@ int ARGBToNV21(const uint8_t* src_argb,
MergeUVRow_ = MergeUVRow_MSA;
}
}
#endif
#if defined(HAS_MERGEUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
MergeUVRow_ = MergeUVRow_Any_LSX;
if (IS_ALIGNED(halfwidth, 16)) {
MergeUVRow_ = MergeUVRow_LSX;
}
}
#endif
{
// Allocate a rows of uv.
@ -768,6 +784,14 @@ int ABGRToNV12(const uint8_t* src_abgr,
MergeUVRow_ = MergeUVRow_MSA;
}
}
#endif
#if defined(HAS_MERGEUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
MergeUVRow_ = MergeUVRow_Any_LSX;
if (IS_ALIGNED(halfwidth, 16)) {
MergeUVRow_ = MergeUVRow_LSX;
}
}
#endif
{
// Allocate a rows of uv.
@ -932,6 +956,14 @@ int ABGRToNV21(const uint8_t* src_abgr,
MergeUVRow_ = MergeUVRow_MSA;
}
}
#endif
#if defined(HAS_MERGEUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
MergeUVRow_ = MergeUVRow_Any_LSX;
if (IS_ALIGNED(halfwidth, 16)) {
MergeUVRow_ = MergeUVRow_LSX;
}
}
#endif
{
// Allocate a rows of uv.
@ -2117,6 +2149,16 @@ int ARGBToJ420(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTOYJROW_LSX) && defined(HAS_ARGBTOUVJROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToYJRow = ARGBToYJRow_Any_LSX;
ARGBToUVJRow = ARGBToUVJRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
ARGBToYJRow = ARGBToYJRow_LSX;
ARGBToUVJRow = ARGBToUVJRow_LSX;
}
}
#endif
for (y = 0; y < height - 1; y += 2) {
ARGBToUVJRow(src_argb, src_stride_argb, dst_u, dst_v, width);
@ -2232,6 +2274,16 @@ int ARGBToJ422(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTOYJROW_LSX) && defined(HAS_ARGBTOUVJROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToYJRow = ARGBToYJRow_Any_LSX;
ARGBToUVJRow = ARGBToUVJRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
ARGBToYJRow = ARGBToYJRow_LSX;
ARGBToUVJRow = ARGBToUVJRow_LSX;
}
}
#endif
for (y = 0; y < height; ++y) {
ARGBToUVJRow(src_argb, 0, dst_u, dst_v, width);

141
source/planar_functions.cc
View File

@ -466,6 +466,14 @@ void SplitUVPlane(const uint8_t* src_uv,
}
}
#endif
#if defined(HAS_SPLITUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
SplitUVRow = SplitUVRow_Any_LSX;
if (IS_ALIGNED(width, 32)) {
SplitUVRow = SplitUVRow_LSX;
}
}
#endif
for (y = 0; y < height; ++y) {
// Copy a row of UV.
@ -541,6 +549,14 @@ void MergeUVPlane(const uint8_t* src_u,
}
}
#endif
#if defined(HAS_MERGEUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
MergeUVRow = MergeUVRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
MergeUVRow = MergeUVRow_LSX;
}
}
#endif
for (y = 0; y < height; ++y) {
// Merge a row of U and V into a row of UV.
@ -2321,6 +2337,11 @@ ARGBBlendRow GetARGBBlend() {
if (TestCpuFlag(kCpuHasMSA)) {
ARGBBlendRow = ARGBBlendRow_MSA;
}
#endif
#if defined(HAS_ARGBBLENDROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBBlendRow = ARGBBlendRow_LSX;
}
#endif
return ARGBBlendRow;
}
@ -2904,6 +2925,14 @@ int RAWToRGB24(const uint8_t* src_raw,
}
}
#endif
#if defined(HAS_RAWTORGB24ROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
RAWToRGB24Row = RAWToRGB24Row_Any_LSX;
if (IS_ALIGNED(width, 16)) {
RAWToRGB24Row = RAWToRGB24Row_LSX;
}
}
#endif
for (y = 0; y < height; ++y) {
RAWToRGB24Row(src_raw, dst_rgb24, width);
@ -2958,6 +2987,14 @@ void SetPlane(uint8_t* dst_y,
SetRow = SetRow_MSA;
}
#endif
#if defined(HAS_SETROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
SetRow = SetRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
SetRow = SetRow_LSX;
}
}
#endif
// Set plane
for (y = 0; y < height; ++y) {
@ -3055,6 +3092,14 @@ int ARGBRect(uint8_t* dst_argb,
}
}
#endif
#if defined(HAS_ARGBSETROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBSetRow = ARGBSetRow_Any_LSX;
if (IS_ALIGNED(width, 4)) {
ARGBSetRow = ARGBSetRow_LSX;
}
}
#endif
// Set plane
for (y = 0; y < height; ++y) {
@ -3422,6 +3467,11 @@ int ARGBColorMatrix(const uint8_t* src_argb,
if (TestCpuFlag(kCpuHasMSA) && IS_ALIGNED(width, 8)) {
ARGBColorMatrixRow = ARGBColorMatrixRow_MSA;
}
#endif
#if defined(HAS_ARGBCOLORMATRIXROW_LSX)
if (TestCpuFlag(kCpuHasLSX) && IS_ALIGNED(width, 8)) {
ARGBColorMatrixRow = ARGBColorMatrixRow_LSX;
}
#endif
for (y = 0; y < height; ++y) {
ARGBColorMatrixRow(src_argb, dst_argb, matrix_argb, width);
@ -3587,6 +3637,11 @@ int ARGBQuantize(uint8_t* dst_argb,
if (TestCpuFlag(kCpuHasMSA) && IS_ALIGNED(width, 8)) {
ARGBQuantizeRow = ARGBQuantizeRow_MSA;
}
#endif
#if defined(HAS_ARGBQUANTIZEROW_LSX)
if (TestCpuFlag(kCpuHasLSX) && IS_ALIGNED(width, 8)) {
ARGBQuantizeRow = ARGBQuantizeRow_LSX;
}
#endif
for (y = 0; y < height; ++y) {
ARGBQuantizeRow(dst, scale, interval_size, interval_offset, width);
@ -3881,6 +3936,14 @@ int InterpolatePlane(const uint8_t* src0,
}
}
#endif
#if defined(HAS_INTERPOLATEROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
InterpolateRow = InterpolateRow_Any_LSX;
if (IS_ALIGNED(width, 32)) {
InterpolateRow = InterpolateRow_LSX;
}
}
#endif
for (y = 0; y < height; ++y) {
InterpolateRow(dst, src0, src1 - src0, width, interpolation);
@ -4243,6 +4306,14 @@ static int ARGBSobelize(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTOYJROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToYJRow = ARGBToYJRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
ARGBToYJRow = ARGBToYJRow_LSX;
}
}
#endif
#if defined(HAS_SOBELYROW_SSE2)
if (TestCpuFlag(kCpuHasSSE2)) {
@ -4373,6 +4444,14 @@ int ARGBSobel(const uint8_t* src_argb,
SobelRow = SobelRow_MSA;
}
}
#endif
#if defined(HAS_SOBELROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
SobelRow = SobelRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
SobelRow = SobelRow_LSX;
}
}
#endif
return ARGBSobelize(src_argb, src_stride_argb, dst_argb, dst_stride_argb,
width, height, SobelRow);
@ -4419,6 +4498,14 @@ int ARGBSobelToPlane(const uint8_t* src_argb,
SobelToPlaneRow = SobelToPlaneRow_MSA;
}
}
#endif
#if defined(HAS_SOBELTOPLANEROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
SobelToPlaneRow = SobelToPlaneRow_Any_LSX;
if (IS_ALIGNED(width, 32)) {
SobelToPlaneRow = SobelToPlaneRow_LSX;
}
}
#endif
return ARGBSobelize(src_argb, src_stride_argb, dst_y, dst_stride_y, width,
height, SobelToPlaneRow);
@ -4466,6 +4553,14 @@ int ARGBSobelXY(const uint8_t* src_argb,
SobelXYRow = SobelXYRow_MSA;
}
}
#endif
#if defined(HAS_SOBELXYROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
SobelXYRow = SobelXYRow_Any_LSX;
if (IS_ALIGNED(width, 16)) {
SobelXYRow = SobelXYRow_LSX;
}
}
#endif
return ARGBSobelize(src_argb, src_stride_argb, dst_argb, dst_stride_argb,
width, height, SobelXYRow);
@ -4590,6 +4685,14 @@ int HalfFloatPlane(const uint16_t* src_y,
}
}
#endif
#if defined(HAS_HALFFLOATROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
HalfFloatRow = HalfFloatRow_Any_LSX;
if (IS_ALIGNED(width, 32)) {
HalfFloatRow = HalfFloatRow_LSX;
}
}
#endif
for (y = 0; y < height; ++y) {
HalfFloatRow(src_y, dst_y, scale, width);
@ -4776,6 +4879,12 @@ int ARGBExtractAlpha(const uint8_t* src_argb,
: ARGBExtractAlphaRow_Any_MSA;
}
#endif
#if defined(HAS_ARGBEXTRACTALPHAROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBExtractAlphaRow = IS_ALIGNED(width, 16) ? ARGBExtractAlphaRow_LSX
: ARGBExtractAlphaRow_Any_LSX;
}
#endif
for (int y = 0; y < height; ++y) {
ARGBExtractAlphaRow(src_argb, dst_a, width);
@ -4912,6 +5021,14 @@ int YUY2ToNV12(const uint8_t* src_yuy2,
}
}
#endif
#if defined(HAS_SPLITUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
SplitUVRow = SplitUVRow_Any_LSX;
if (IS_ALIGNED(width, 32)) {
SplitUVRow = SplitUVRow_LSX;
}
}
#endif
#if defined(HAS_INTERPOLATEROW_SSSE3)
if (TestCpuFlag(kCpuHasSSSE3)) {
InterpolateRow = InterpolateRow_Any_SSSE3;
@ -4952,6 +5069,14 @@ int YUY2ToNV12(const uint8_t* src_yuy2,
}
}
#endif
#if defined(HAS_INTERPOLATEROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
InterpolateRow = InterpolateRow_Any_LSX;
if (IS_ALIGNED(width, 32)) {
InterpolateRow = InterpolateRow_LSX;
}
}
#endif
{
int awidth = halfwidth * 2;
@ -5044,6 +5169,14 @@ int UYVYToNV12(const uint8_t* src_uyvy,
}
}
#endif
#if defined(HAS_SPLITUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
SplitUVRow = SplitUVRow_Any_LSX;
if (IS_ALIGNED(width, 32)) {
SplitUVRow = SplitUVRow_LSX;
}
}
#endif
#if defined(HAS_INTERPOLATEROW_SSSE3)
if (TestCpuFlag(kCpuHasSSSE3)) {
InterpolateRow = InterpolateRow_Any_SSSE3;
@ -5084,6 +5217,14 @@ int UYVYToNV12(const uint8_t* src_uyvy,
}
}
#endif
#if defined(HAS_INTERPOLATEROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
InterpolateRow = InterpolateRow_Any_LSX;
if (IS_ALIGNED(width, 32)) {
InterpolateRow = InterpolateRow_LSX;
}
}
#endif
{
int awidth = halfwidth * 2;

5
source/rotate.cc
View File

@ -366,6 +366,11 @@ void SplitRotateUV180(const uint8_t* src,
MirrorSplitUVRow = MirrorSplitUVRow_MSA;
}
#endif
#if defined(HAS_MIRRORSPLITUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX) && IS_ALIGNED(width, 32)) {
MirrorSplitUVRow = MirrorSplitUVRow_LSX;
}
#endif
dst_a += dst_stride_a * (height - 1);
dst_b += dst_stride_b * (height - 1);

123
source/row_any.cc
View File

@ -436,6 +436,9 @@ ANY31C(I422ToRGB565Row_Any_LASX, I422ToRGB565Row_LASX, 1, 0, 2, 31)
ANY31C(I422ToARGB4444Row_Any_LASX, I422ToARGB4444Row_LASX, 1, 0, 2, 31)
ANY31C(I422ToARGB1555Row_Any_LASX, I422ToARGB1555Row_LASX, 1, 0, 2, 31)
#endif
#ifdef HAS_I444TOARGBROW_LSX
ANY31C(I444ToARGBRow_Any_LSX, I444ToARGBRow_LSX, 0, 0, 4, 15)
#endif
#undef ANY31C
// Any 3 planes of 16 bit to 1 with yuvconstants
@ -598,6 +601,9 @@ ANY21(MergeUVRow_Any_MSA, MergeUVRow_MSA, 0, 1, 1, 2, 15)
#ifdef HAS_MERGEUVROW_MMI
ANY21(MergeUVRow_Any_MMI, MergeUVRow_MMI, 0, 1, 1, 2, 7)
#endif
#ifdef HAS_MERGEUVROW_LSX
ANY21(MergeUVRow_Any_LSX, MergeUVRow_LSX, 0, 1, 1, 2, 15)
#endif
#ifdef HAS_NV21TOYUV24ROW_NEON
ANY21(NV21ToYUV24Row_Any_NEON, NV21ToYUV24Row_NEON, 1, 1, 2, 3, 15)
#endif
@ -674,6 +680,9 @@ ANY21(SobelRow_Any_MSA, SobelRow_MSA, 0, 1, 1, 4, 15)
#ifdef HAS_SOBELROW_MMI
ANY21(SobelRow_Any_MMI, SobelRow_MMI, 0, 1, 1, 4, 7)
#endif
#ifdef HAS_SOBELROW_LSX
ANY21(SobelRow_Any_LSX, SobelRow_LSX, 0, 1, 1, 4, 15)
#endif
#ifdef HAS_SOBELTOPLANEROW_SSE2
ANY21(SobelToPlaneRow_Any_SSE2, SobelToPlaneRow_SSE2, 0, 1, 1, 1, 15)
#endif
@ -686,6 +695,9 @@ ANY21(SobelToPlaneRow_Any_MSA, SobelToPlaneRow_MSA, 0, 1, 1, 1, 31)
#ifdef HAS_SOBELTOPLANEROW_MMI
ANY21(SobelToPlaneRow_Any_MMI, SobelToPlaneRow_MMI, 0, 1, 1, 1, 7)
#endif
#ifdef HAS_SOBELTOPLANEROW_LSX
ANY21(SobelToPlaneRow_Any_LSX, SobelToPlaneRow_LSX, 0, 1, 1, 1, 31)
#endif
#ifdef HAS_SOBELXYROW_SSE2
ANY21(SobelXYRow_Any_SSE2, SobelXYRow_SSE2, 0, 1, 1, 4, 15)
#endif
@ -698,6 +710,9 @@ ANY21(SobelXYRow_Any_MSA, SobelXYRow_MSA, 0, 1, 1, 4, 15)
#ifdef HAS_SOBELXYROW_MMI
ANY21(SobelXYRow_Any_MMI, SobelXYRow_MMI, 0, 1, 1, 4, 7)
#endif
#ifdef HAS_SOBELXYROW_LSX
ANY21(SobelXYRow_Any_LSX, SobelXYRow_LSX, 0, 1, 1, 4, 15)
#endif
#undef ANY21
// Any 2 planes to 1 with yuvconstants
@ -734,6 +749,9 @@ ANY21C(NV12ToARGBRow_Any_MSA, NV12ToARGBRow_MSA, 1, 1, 2, 4, 7)
#ifdef HAS_NV12TOARGBROW_MMI
ANY21C(NV12ToARGBRow_Any_MMI, NV12ToARGBRow_MMI, 1, 1, 2, 4, 7)
#endif
#ifdef HAS_NV12TOARGBROW_LSX
ANY21C(NV12ToARGBRow_Any_LSX, NV12ToARGBRow_LSX, 1, 1, 2, 4, 7)
#endif
#ifdef HAS_NV21TOARGBROW_SSSE3
ANY21C(NV21ToARGBRow_Any_SSSE3, NV21ToARGBRow_SSSE3, 1, 1, 2, 4, 7)
#endif
@ -749,6 +767,9 @@ ANY21C(NV21ToARGBRow_Any_MSA, NV21ToARGBRow_MSA, 1, 1, 2, 4, 7)
#ifdef HAS_NV21TOARGBROW_MMI
ANY21C(NV21ToARGBRow_Any_MMI, NV21ToARGBRow_MMI, 1, 1, 2, 4, 7)
#endif
#ifdef HAS_NV21TOARGBROW_LSX
ANY21C(NV21ToARGBRow_Any_LSX, NV21ToARGBRow_LSX, 1, 1, 2, 4, 7)
#endif
#ifdef HAS_NV12TORGB24ROW_NEON
ANY21C(NV12ToRGB24Row_Any_NEON, NV12ToRGB24Row_NEON, 1, 1, 2, 3, 7)
#endif
@ -788,6 +809,9 @@ ANY21C(NV12ToRGB565Row_Any_MSA, NV12ToRGB565Row_MSA, 1, 1, 2, 2, 7)
#ifdef HAS_NV12TORGB565ROW_MMI
ANY21C(NV12ToRGB565Row_Any_MMI, NV12ToRGB565Row_MMI, 1, 1, 2, 2, 7)
#endif
#ifdef HAS_NV12TORGB565ROW_LSX
ANY21C(NV12ToRGB565Row_Any_LSX, NV12ToRGB565Row_LSX, 1, 1, 2, 2, 7)
#endif
#undef ANY21C
// Any 2 planes of 16 bit to 1 with yuvconstants
@ -979,6 +1003,9 @@ ANY11(ARGBToRGB565Row_Any_LASX, ARGBToRGB565Row_LASX, 0, 4, 2, 15)
ANY11(ARGBToARGB1555Row_Any_LASX, ARGBToARGB1555Row_LASX, 0, 4, 2, 15)
ANY11(ARGBToARGB4444Row_Any_LASX, ARGBToARGB4444Row_LASX, 0, 4, 2, 15)
#endif
#if defined(HAS_J400TOARGBROW_LSX)
ANY11(J400ToARGBRow_Any_LSX, J400ToARGBRow_LSX, 0, 1, 4, 15)
#endif
#if defined(HAS_RAWTORGB24ROW_NEON)
ANY11(RAWToRGB24Row_Any_NEON, RAWToRGB24Row_NEON, 0, 3, 3, 7)
#endif
@ -988,6 +1015,9 @@ ANY11(RAWToRGB24Row_Any_MSA, RAWToRGB24Row_MSA, 0, 3, 3, 15)
#if defined(HAS_RAWTORGB24ROW_MMI)
ANY11(RAWToRGB24Row_Any_MMI, RAWToRGB24Row_MMI, 0, 3, 3, 3)
#endif
#if defined(HAS_RAWTORGB24ROW_LSX)
ANY11(RAWToRGB24Row_Any_LSX, RAWToRGB24Row_LSX, 0, 3, 3, 15)
#endif
#ifdef HAS_ARGBTOYROW_AVX2
ANY11(ARGBToYRow_Any_AVX2, ARGBToYRow_AVX2, 0, 4, 1, 31)
#endif
@ -1048,6 +1078,9 @@ ANY11(ARGBToYJRow_Any_MSA, ARGBToYJRow_MSA, 0, 4, 1, 15)
#ifdef HAS_ARGBTOYJROW_MMI
ANY11(ARGBToYJRow_Any_MMI, ARGBToYJRow_MMI, 0, 4, 1, 7)
#endif
#ifdef HAS_ARGBTOYJROW_LSX
ANY11(ARGBToYJRow_Any_LSX, ARGBToYJRow_LSX, 0, 4, 1, 15)
#endif
#ifdef HAS_BGRATOYROW_NEON
ANY11(BGRAToYRow_Any_NEON, BGRAToYRow_NEON, 0, 4, 1, 7)
#endif
@ -1057,6 +1090,9 @@ ANY11(BGRAToYRow_Any_MSA, BGRAToYRow_MSA, 0, 4, 1, 15)
#ifdef HAS_BGRATOYROW_MMI
ANY11(BGRAToYRow_Any_MMI, BGRAToYRow_MMI, 0, 4, 1, 7)
#endif
#ifdef HAS_BGRATOYROW_LSX
ANY11(BGRAToYRow_Any_LSX, BGRAToYRow_LSX, 0, 4, 1, 15)
#endif
#ifdef HAS_ABGRTOYROW_NEON
ANY11(ABGRToYRow_Any_NEON, ABGRToYRow_NEON, 0, 4, 1, 7)
#endif
@ -1066,6 +1102,9 @@ ANY11(ABGRToYRow_Any_MSA, ABGRToYRow_MSA, 0, 4, 1, 7)
#ifdef HAS_ABGRTOYROW_MMI
ANY11(ABGRToYRow_Any_MMI, ABGRToYRow_MMI, 0, 4, 1, 7)
#endif
#ifdef HAS_ABGRTOYROW_LSX
ANY11(ABGRToYRow_Any_LSX, ABGRToYRow_LSX, 0, 4, 1, 15)
#endif
#ifdef HAS_RGBATOYROW_NEON
ANY11(RGBAToYRow_Any_NEON, RGBAToYRow_NEON, 0, 4, 1, 7)
#endif
@ -1075,6 +1114,9 @@ ANY11(RGBAToYRow_Any_MSA, RGBAToYRow_MSA, 0, 4, 1, 15)
#ifdef HAS_RGBATOYROW_MMI
ANY11(RGBAToYRow_Any_MMI, RGBAToYRow_MMI, 0, 4, 1, 7)
#endif
#ifdef HAS_RGBATOYROW_LSX
ANY11(RGBAToYRow_Any_LSX, RGBAToYRow_LSX, 0, 4, 1, 15)
#endif
#ifdef HAS_RGB24TOYROW_NEON
ANY11(RGB24ToYRow_Any_NEON, RGB24ToYRow_NEON, 0, 3, 1, 7)
#endif
@ -1093,6 +1135,9 @@ ANY11(RGB24ToYRow_Any_MSA, RGB24ToYRow_MSA, 0, 3, 1, 15)
#ifdef HAS_RGB24TOYROW_MMI
ANY11(RGB24ToYRow_Any_MMI, RGB24ToYRow_MMI, 0, 3, 1, 7)
#endif
#ifdef HAS_RGB24TOYROW_LSX
ANY11(RGB24ToYRow_Any_LSX, RGB24ToYRow_LSX, 0, 3, 1, 15)
#endif
#ifdef HAS_RAWTOYROW_NEON
ANY11(RAWToYRow_Any_NEON, RAWToYRow_NEON, 0, 3, 1, 7)
#endif
@ -1111,6 +1156,9 @@ ANY11(RAWToYRow_Any_MSA, RAWToYRow_MSA, 0, 3, 1, 15)
#ifdef HAS_RAWTOYROW_MMI
ANY11(RAWToYRow_Any_MMI, RAWToYRow_MMI, 0, 3, 1, 7)
#endif
#ifdef HAS_RAWTOYROW_LSX
ANY11(RAWToYRow_Any_LSX, RAWToYRow_LSX, 0, 3, 1, 15)
#endif
#ifdef HAS_RGB565TOYROW_NEON
ANY11(RGB565ToYRow_Any_NEON, RGB565ToYRow_NEON, 0, 2, 1, 7)
#endif
@ -1120,6 +1168,9 @@ ANY11(RGB565ToYRow_Any_MSA, RGB565ToYRow_MSA, 0, 2, 1, 15)
#ifdef HAS_RGB565TOYROW_MMI
ANY11(RGB565ToYRow_Any_MMI, RGB565ToYRow_MMI, 0, 2, 1, 7)
#endif
#ifdef HAS_RGB565TOYROW_LSX
ANY11(RGB565ToYRow_Any_LSX, RGB565ToYRow_LSX, 0, 2, 1, 15)
#endif
#ifdef HAS_ARGB1555TOYROW_NEON
ANY11(ARGB1555ToYRow_Any_NEON, ARGB1555ToYRow_NEON, 0, 2, 1, 7)
#endif
@ -1129,6 +1180,9 @@ ANY11(ARGB1555ToYRow_Any_MSA, ARGB1555ToYRow_MSA, 0, 2, 1, 15)
#ifdef HAS_ARGB1555TOYROW_MMI
ANY11(ARGB1555ToYRow_Any_MMI, ARGB1555ToYRow_MMI, 0, 2, 1, 7)
#endif
#ifdef HAS_ARGB1555TOYROW_LSX
ANY11(ARGB1555ToYRow_Any_LSX, ARGB1555ToYRow_LSX, 0, 2, 1, 15)
#endif
#ifdef HAS_ARGB4444TOYROW_NEON
ANY11(ARGB4444ToYRow_Any_NEON, ARGB4444ToYRow_NEON, 0, 2, 1, 7)
#endif
@ -1180,6 +1234,9 @@ ANY11(RGB24ToARGBRow_Any_MSA, RGB24ToARGBRow_MSA, 0, 3, 4, 15)
#ifdef HAS_RGB24TOARGBROW_MMI
ANY11(RGB24ToARGBRow_Any_MMI, RGB24ToARGBRow_MMI, 0, 3, 4, 3)
#endif
#ifdef HAS_RGB24TOARGBROW_LSX
ANY11(RGB24ToARGBRow_Any_LSX, RGB24ToARGBRow_LSX, 0, 3, 4, 15)
#endif
#ifdef HAS_RAWTOARGBROW_NEON
ANY11(RAWToARGBRow_Any_NEON, RAWToARGBRow_NEON, 0, 3, 4, 7)
#endif
@ -1192,6 +1249,9 @@ ANY11(RAWToARGBRow_Any_MSA, RAWToARGBRow_MSA, 0, 3, 4, 15)
#ifdef HAS_RAWTOARGBROW_MMI
ANY11(RAWToARGBRow_Any_MMI, RAWToARGBRow_MMI, 0, 3, 4, 3)
#endif
#ifdef HAS_RAWTOARGBROW_LSX
ANY11(RAWToARGBRow_Any_LSX, RAWToARGBRow_LSX, 0, 3, 4, 15)
#endif
#ifdef HAS_RGB565TOARGBROW_NEON
ANY11(RGB565ToARGBRow_Any_NEON, RGB565ToARGBRow_NEON, 0, 2, 4, 7)
#endif
@ -1201,6 +1261,9 @@ ANY11(RGB565ToARGBRow_Any_MSA, RGB565ToARGBRow_MSA, 0, 2, 4, 15)
#ifdef HAS_RGB565TOARGBROW_MMI
ANY11(RGB565ToARGBRow_Any_MMI, RGB565ToARGBRow_MMI, 0, 2, 4, 3)
#endif
#ifdef HAS_RGB565TOARGBROW_LSX
ANY11(RGB565ToARGBRow_Any_LSX, RGB565ToARGBRow_LSX, 0, 2, 4, 15)
#endif
#ifdef HAS_ARGB1555TOARGBROW_NEON
ANY11(ARGB1555ToARGBRow_Any_NEON, ARGB1555ToARGBRow_NEON, 0, 2, 4, 7)
#endif
@ -1210,6 +1273,9 @@ ANY11(ARGB1555ToARGBRow_Any_MSA, ARGB1555ToARGBRow_MSA, 0, 2, 4, 15)
#ifdef HAS_ARGB1555TOARGBROW_MMI
ANY11(ARGB1555ToARGBRow_Any_MMI, ARGB1555ToARGBRow_MMI, 0, 2, 4, 3)
#endif
#ifdef HAS_ARGB1555TOARGBROW_LSX
ANY11(ARGB1555ToARGBRow_Any_LSX, ARGB1555ToARGBRow_LSX, 0, 2, 4, 15)
#endif
#ifdef HAS_ARGB4444TOARGBROW_NEON
ANY11(ARGB4444ToARGBRow_Any_NEON, ARGB4444ToARGBRow_NEON, 0, 2, 4, 7)
#endif
@ -1219,6 +1285,9 @@ ANY11(ARGB4444ToARGBRow_Any_MSA, ARGB4444ToARGBRow_MSA, 0, 2, 4, 15)
#ifdef HAS_ARGB4444TOARGBROW_MMI
ANY11(ARGB4444ToARGBRow_Any_MMI, ARGB4444ToARGBRow_MMI, 0, 2, 4, 3)
#endif
#ifdef HAS_ARGB4444TOARGBROW_LSX
ANY11(ARGB4444ToARGBRow_Any_LSX, ARGB4444ToARGBRow_LSX, 0, 2, 4, 15)
#endif
#ifdef HAS_ARGBATTENUATEROW_SSSE3
ANY11(ARGBAttenuateRow_Any_SSSE3, ARGBAttenuateRow_SSSE3, 0, 4, 4, 3)
#endif
@ -1258,6 +1327,9 @@ ANY11(ARGBExtractAlphaRow_Any_MSA, ARGBExtractAlphaRow_MSA, 0, 4, 1, 15)
#ifdef HAS_ARGBEXTRACTALPHAROW_MMI
ANY11(ARGBExtractAlphaRow_Any_MMI, ARGBExtractAlphaRow_MMI, 0, 4, 1, 7)
#endif
#ifdef HAS_ARGBEXTRACTALPHAROW_LSX
ANY11(ARGBExtractAlphaRow_Any_LSX, ARGBExtractAlphaRow_LSX, 0, 4, 1, 15)
#endif
#undef ANY11
// Any 1 to 1 blended. Destination is read, modify, write.
@ -1351,6 +1423,14 @@ ANY11P(I400ToARGBRow_Any_MMI,
4,
7)
#endif
#if defined(HAS_I400TOARGBROW_LSX)
ANY11P(I400ToARGBRow_Any_LSX,
I400ToARGBRow_LSX,
const struct YuvConstants*,
1,
4,
15)
#endif
#if defined(HAS_ARGBTORGB565DITHERROW_SSE2)
ANY11P(ARGBToRGB565DitherRow_Any_SSE2,
@ -1613,6 +1693,9 @@ ANY11P16(HalfFloatRow_Any_MSA, HalfFloatRow_MSA, uint16_t, uint16_t, 2, 2, 31)
#ifdef HAS_BYTETOFLOATROW_NEON
ANY11P16(ByteToFloatRow_Any_NEON, ByteToFloatRow_NEON, uint8_t, float, 1, 3, 7)
#endif
#ifdef HAS_HALFFLOATROW_LSX
ANY11P16(HalfFloatRow_Any_LSX, HalfFloatRow_LSX, uint16_t, uint16_t, 2, 2, 31)
#endif
#undef ANY11P16
// Any 1 to 1 with yuvconstants
@ -1650,6 +1733,10 @@ ANY11C(UYVYToARGBRow_Any_MSA, UYVYToARGBRow_MSA, 1, 4, 4, 7)
ANY11C(YUY2ToARGBRow_Any_MMI, YUY2ToARGBRow_MMI, 1, 4, 4, 7)
ANY11C(UYVYToARGBRow_Any_MMI, UYVYToARGBRow_MMI, 1, 4, 4, 7)
#endif
#if defined(HAS_YUY2TOARGBROW_LSX)
ANY11C(YUY2ToARGBRow_Any_LSX, YUY2ToARGBRow_LSX, 1, 4, 4, 7)
ANY11C(UYVYToARGBRow_Any_LSX, UYVYToARGBRow_LSX, 1, 4, 4, 7)
#endif
#undef ANY11C
// Any 1 to 1 interpolate. Takes 2 rows of source via stride.
@ -1684,6 +1771,9 @@ ANY11I(InterpolateRow_Any_MSA, InterpolateRow_MSA, 1, 1, 31)
#ifdef HAS_INTERPOLATEROW_MMI
ANY11I(InterpolateRow_Any_MMI, InterpolateRow_MMI, 1, 1, 7)
#endif
#ifdef HAS_INTERPOLATEROW_LSX
ANY11I(InterpolateRow_Any_LSX, InterpolateRow_LSX, 1, 1, 31)
#endif
#undef ANY11I
// Any 1 to 1 mirror.
@ -1780,6 +1870,9 @@ ANY1(SetRow_Any_X86, SetRow_X86, uint8_t, 1, 3)
#ifdef HAS_SETROW_NEON
ANY1(SetRow_Any_NEON, SetRow_NEON, uint8_t, 1, 15)
#endif
#ifdef HAS_SETROW_LSX
ANY1(SetRow_Any_LSX, SetRow_LSX, uint8_t, 1, 15)
#endif
#ifdef HAS_ARGBSETROW_NEON
ANY1(ARGBSetRow_Any_NEON, ARGBSetRow_NEON, uint32_t, 4, 3)
#endif
@ -1789,6 +1882,9 @@ ANY1(ARGBSetRow_Any_MSA, ARGBSetRow_MSA, uint32_t, 4, 3)
#ifdef HAS_ARGBSETROW_MMI
ANY1(ARGBSetRow_Any_MMI, ARGBSetRow_MMI, uint32_t, 4, 3)
#endif
#ifdef HAS_ARGBSETROW_LSX
ANY1(ARGBSetRow_Any_LSX, ARGBSetRow_LSX, uint32_t, 4, 3)
#endif
#undef ANY1
// Any 1 to 2. Outputs UV planes.
@ -1823,6 +1919,9 @@ ANY12(SplitUVRow_Any_MSA, SplitUVRow_MSA, 0, 2, 0, 31)
#ifdef HAS_SPLITUVROW_MMI
ANY12(SplitUVRow_Any_MMI, SplitUVRow_MMI, 0, 2, 0, 7)
#endif
#ifdef HAS_SPLITUVROW_LSX
ANY12(SplitUVRow_Any_LSX, SplitUVRow_LSX, 0, 2, 0, 31)
#endif
#ifdef HAS_ARGBTOUV444ROW_SSSE3
ANY12(ARGBToUV444Row_Any_SSSE3, ARGBToUV444Row_SSSE3, 0, 4, 0, 15)
#endif
@ -2026,12 +2125,18 @@ ANY12S(ARGBToUVJRow_Any_MSA, ARGBToUVJRow_MSA, 0, 4, 31)
#ifdef HAS_ARGBTOUVJROW_MMI
ANY12S(ARGBToUVJRow_Any_MMI, ARGBToUVJRow_MMI, 0, 4, 15)
#endif
#ifdef HAS_ARGBTOUVJROW_LSX
ANY12S(ARGBToUVJRow_Any_LSX, ARGBToUVJRow_LSX, 0, 4, 15)
#endif
#ifdef HAS_BGRATOUVROW_NEON
ANY12S(BGRAToUVRow_Any_NEON, BGRAToUVRow_NEON, 0, 4, 15)
#endif
#ifdef HAS_BGRATOUVROW_MSA
ANY12S(BGRAToUVRow_Any_MSA, BGRAToUVRow_MSA, 0, 4, 15)
#endif
#ifdef HAS_BGRATOUVROW_LSX
ANY12S(BGRAToUVRow_Any_LSX, BGRAToUVRow_LSX, 0, 4, 15)
#endif
#ifdef HAS_BGRATOUVROW_MMI
ANY12S(BGRAToUVRow_Any_MMI, BGRAToUVRow_MMI, 0, 4, 15)
#endif
@ -2044,6 +2149,9 @@ ANY12S(ABGRToUVRow_Any_MSA, ABGRToUVRow_MSA, 0, 4, 15)
#ifdef HAS_ABGRTOUVROW_MMI
ANY12S(ABGRToUVRow_Any_MMI, ABGRToUVRow_MMI, 0, 4, 15)
#endif
#ifdef HAS_ABGRTOUVROW_LSX
ANY12S(ABGRToUVRow_Any_LSX, ABGRToUVRow_LSX, 0, 4, 15)
#endif
#ifdef HAS_RGBATOUVROW_NEON
ANY12S(RGBAToUVRow_Any_NEON, RGBAToUVRow_NEON, 0, 4, 15)
#endif
@ -2053,6 +2161,9 @@ ANY12S(RGBAToUVRow_Any_MSA, RGBAToUVRow_MSA, 0, 4, 15)
#ifdef HAS_RGBATOUVROW_MMI
ANY12S(RGBAToUVRow_Any_MMI, RGBAToUVRow_MMI, 0, 4, 15)
#endif
#ifdef HAS_RGBATOUVROW_LSX
ANY12S(RGBAToUVRow_Any_LSX, RGBAToUVRow_LSX, 0, 4, 15)
#endif
#ifdef HAS_RGB24TOUVROW_NEON
ANY12S(RGB24ToUVRow_Any_NEON, RGB24ToUVRow_NEON, 0, 3, 15)
#endif
@ -2065,6 +2176,9 @@ ANY12S(RGB24ToUVRow_Any_MSA, RGB24ToUVRow_MSA, 0, 3, 15)
#ifdef HAS_RGB24TOUVROW_MMI
ANY12S(RGB24ToUVRow_Any_MMI, RGB24ToUVRow_MMI, 0, 3, 15)
#endif
#ifdef HAS_RGB24TOUVROW_LSX
ANY12S(RGB24ToUVRow_Any_LSX, RGB24ToUVRow_LSX, 0, 3, 15)
#endif
#ifdef HAS_RAWTOUVROW_NEON
ANY12S(RAWToUVRow_Any_NEON, RAWToUVRow_NEON, 0, 3, 15)
#endif
@ -2077,6 +2191,9 @@ ANY12S(RAWToUVRow_Any_MSA, RAWToUVRow_MSA, 0, 3, 15)
#ifdef HAS_RAWTOUVROW_MMI
ANY12S(RAWToUVRow_Any_MMI, RAWToUVRow_MMI, 0, 3, 15)
#endif
#ifdef HAS_RAWTOUVROW_LSX
ANY12S(RAWToUVRow_Any_LSX, RAWToUVRow_LSX, 0, 3, 15)
#endif
#ifdef HAS_RGB565TOUVROW_NEON
ANY12S(RGB565ToUVRow_Any_NEON, RGB565ToUVRow_NEON, 0, 2, 15)
#endif
@ -2086,6 +2203,9 @@ ANY12S(RGB565ToUVRow_Any_MSA, RGB565ToUVRow_MSA, 0, 2, 15)
#ifdef HAS_RGB565TOUVROW_MMI
ANY12S(RGB565ToUVRow_Any_MMI, RGB565ToUVRow_MMI, 0, 2, 15)
#endif
#ifdef HAS_RGB565TOUVROW_LSX
ANY12S(RGB565ToUVRow_Any_LSX, RGB565ToUVRow_LSX, 0, 2, 15)
#endif
#ifdef HAS_ARGB1555TOUVROW_NEON
ANY12S(ARGB1555ToUVRow_Any_NEON, ARGB1555ToUVRow_NEON, 0, 2, 15)
#endif
@ -2095,6 +2215,9 @@ ANY12S(ARGB1555ToUVRow_Any_MSA, ARGB1555ToUVRow_MSA, 0, 2, 15)
#ifdef HAS_ARGB1555TOUVROW_MMI
ANY12S(ARGB1555ToUVRow_Any_MMI, ARGB1555ToUVRow_MMI, 0, 2, 15)
#endif
#ifdef HAS_ARGB1555TOUVROW_LSX
ANY12S(ARGB1555ToUVRow_Any_LSX, ARGB1555ToUVRow_LSX, 0, 2, 15)
#endif
#ifdef HAS_ARGB4444TOUVROW_NEON
ANY12S(ARGB4444ToUVRow_Any_NEON, ARGB4444ToUVRow_NEON, 0, 2, 15)
#endif

1817
source/row_lsx.cc Normal file
View File

File diff suppressed because it is too large Load Diff

8
source/scale.cc
View File

@ -1059,6 +1059,14 @@ void ScalePlaneBilinearDown(int src_width,
}
}
#endif
#if defined(HAS_INTERPOLATEROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
InterpolateRow = InterpolateRow_Any_LSX;
if (IS_ALIGNED(src_width, 32)) {
InterpolateRow = InterpolateRow_LSX;
}
}
#endif
#if defined(HAS_SCALEFILTERCOLS_SSSE3)
if (TestCpuFlag(kCpuHasSSSE3) && src_width < 32768) {

24
source/scale_argb.cc
View File

@ -340,6 +340,14 @@ static void ScaleARGBBilinearDown(int src_width,
}
}
#endif
#if defined(HAS_INTERPOLATEROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
InterpolateRow = InterpolateRow_Any_LSX;
if (IS_ALIGNED(clip_src_width, 32)) {
InterpolateRow = InterpolateRow_LSX;
}
}
#endif
#if defined(HAS_SCALEARGBFILTERCOLS_SSSE3)
if (TestCpuFlag(kCpuHasSSSE3) && src_width < 32768) {
ScaleARGBFilterCols = ScaleARGBFilterCols_SSSE3;
@ -451,6 +459,14 @@ static void ScaleARGBBilinearUp(int src_width,
InterpolateRow = InterpolateRow_MSA;
}
}
#endif
#if defined(HAS_INTERPOLATEROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
InterpolateRow = InterpolateRow_Any_LSX;
if (IS_ALIGNED(dst_width, 8)) {
InterpolateRow = InterpolateRow_LSX;
}
}
#endif
if (src_width >= 32768) {
ScaleARGBFilterCols =
@ -687,6 +703,14 @@ static void ScaleYUVToARGBBilinearUp(int src_width,
}
}
#endif
#if defined(HAS_INTERPOLATEROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
InterpolateRow = InterpolateRow_Any_LSX;
if (IS_ALIGNED(dst_width, 8)) {
InterpolateRow = InterpolateRow_LSX;
}
}
#endif
void (*ScaleARGBFilterCols)(uint8_t * dst_argb, const uint8_t* src_argb,
int dst_width, int x, int dx) =

8
source/scale_common.cc
View File

@ -1518,6 +1518,14 @@ void ScalePlaneVertical(int src_height,
InterpolateRow = InterpolateRow_MSA;
}
}
#endif
#if defined(HAS_INTERPOLATEROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
InterpolateRow = InterpolateRow_Any_LSX;
if (IS_ALIGNED(dst_width_bytes, 32)) {
InterpolateRow = InterpolateRow_LSX;
}
}
#endif
for (j = 0; j < dst_height; ++j) {
int yi;

16
source/scale_uv.cc
View File

@ -415,6 +415,14 @@ static void ScaleUVBilinearDown(int src_width,
}
}
#endif
#if defined(HAS_INTERPOLATEROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
InterpolateRow = InterpolateRow_Any_LSX;
if (IS_ALIGNED(clip_src_width, 32)) {
InterpolateRow = InterpolateRow_LSX;
}
}
#endif
#if defined(HAS_SCALEUVFILTERCOLS_SSSE3)
if (TestCpuFlag(kCpuHasSSSE3) && src_width < 32768) {
ScaleUVFilterCols = ScaleUVFilterCols_SSSE3;
@ -528,6 +536,14 @@ static void ScaleUVBilinearUp(int src_width,
InterpolateRow = InterpolateRow_MSA;
}
}
#endif
#if defined(HAS_INTERPOLATEROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
InterpolateRow = InterpolateRow_Any_LSX;
if (IS_ALIGNED(dst_width, 16)) {
InterpolateRow = InterpolateRow_LSX;
}
}
#endif
if (src_width >= 32768) {
ScaleUVFilterCols = filtering ? ScaleUVFilterCols64_C : ScaleUVCols64_C;