diff options
| author | Manojkumar Bhosale <manojkumar.bhosale@imgtec.com> | 2017-02-22 16:06:37 +0530 |
|---|---|---|
| committer | Frank Barchard <fbarchard@google.com> | 2017-02-23 01:42:22 +0000 |
| commit | 45b176d153c61732d81b2df0109bd292e67d2b6b (patch) | |
| tree | 5a410854559b10c0b58d00c827a22d286a35d033 /source/scale_argb.cc | |
| parent | a041b0ae03c0b58ccb6598a846a268c3b2cabf65 (diff) | |
| download | libyuv-45b176d153c61732d81b2df0109bd292e67d2b6b.tar.gz | |
Add MSA optimized Interpolate/MergeUV/Misc functions
BUG=libyuv:634
Change-Id: If8d60bd57f01fe95bc2fd26196466574195cc126
Performance Gain (vs C auto-vectorized)
InterpolateRow_MSA - ~3.3x
InterpolateRow_Any_MSA - ~2.5x
ARGBSetRow_MSA - ~1.0x
ARGBSetRow_Any_MSA - ~1.0x
ARGBToRGB24Row_MSA - ~1.9x
ARGBToRGB24Row_Any_MSA - ~1.6x
MergeUVRow_MSA - ~1.6x
MergeUVRow_Any_MSA - ~1.2x
Performance Gain (vs C non-vectorized)
InterpolateRow_MSA - ~11.3x
InterpolateRow_Any_MSA - ~ 7.9x
ARGBSetRow_MSA - ~ 6.2x
ARGBSetRow_Any_MSA - ~ 4.0x
ARGBToRGB24Row_MSA - ~ 9.9x
ARGBToRGB24Row_Any_MSA - ~ 8.4x
MergeUVRow_MSA - ~12.7x
MergeUVRow_Any_MSA - ~ 8.0x
Change-Id: If8d60bd57f01fe95bc2fd26196466574195cc126
Reviewed-on: https://chromium-review.googlesource.com/445817
Reviewed-by: Frank Barchard <fbarchard@google.com>
Commit-Queue: Frank Barchard <fbarchard@google.com>
Diffstat (limited to 'source/scale_argb.cc')
| -rw-r--r-- | source/scale_argb.cc | 24 |
1 files changed, 24 insertions, 0 deletions
diff --git a/source/scale_argb.cc b/source/scale_argb.cc index dc982d9e..1ea28f0d 100644 --- a/source/scale_argb.cc +++ b/source/scale_argb.cc @@ -315,6 +315,14 @@ static void ScaleARGBBilinearDown(int src_width, } } #endif +#if defined(HAS_INTERPOLATEROW_MSA) + if (TestCpuFlag(kCpuHasMSA)) { + InterpolateRow = InterpolateRow_Any_MSA; + if (IS_ALIGNED(clip_src_width, 32)) { + InterpolateRow = InterpolateRow_MSA; + } + } +#endif #if defined(HAS_SCALEARGBFILTERCOLS_SSSE3) if (TestCpuFlag(kCpuHasSSSE3) && src_width < 32768) { ScaleARGBFilterCols = ScaleARGBFilterCols_SSSE3; @@ -409,6 +417,14 @@ static void ScaleARGBBilinearUp(int src_width, InterpolateRow = InterpolateRow_DSPR2; } #endif +#if defined(HAS_INTERPOLATEROW_MSA) + if (TestCpuFlag(kCpuHasMSA)) { + InterpolateRow = InterpolateRow_Any_MSA; + if (IS_ALIGNED(dst_width, 8)) { + InterpolateRow = InterpolateRow_MSA; + } + } +#endif if (src_width >= 32768) { ScaleARGBFilterCols = filtering ? ScaleARGBFilterCols64_C : ScaleARGBCols64_C; @@ -598,6 +614,14 @@ static void ScaleYUVToARGBBilinearUp(int src_width, InterpolateRow = InterpolateRow_DSPR2; } #endif +#if defined(HAS_INTERPOLATEROW_MSA) + if (TestCpuFlag(kCpuHasMSA)) { + InterpolateRow = InterpolateRow_Any_MSA; + if (IS_ALIGNED(dst_width, 8)) { + InterpolateRow = InterpolateRow_MSA; + } + } +#endif void (*ScaleARGBFilterCols)(uint8 * dst_argb, const uint8* src_argb, int dst_width, int x, int dx) = |