diff options
| author | Manojkumar Bhosale <manojkumar.bhosale@imgtec.com> | 2017-02-01 10:31:28 +0530 |
|---|---|---|
| committer | Manojkumar Bhosale <manojkumar.bhosale@imgtec.com> | 2017-02-01 10:31:28 +0530 |
| commit | 54ce8f23d6d82d671ec05229812842548faf14ef (patch) | |
| tree | 8f4f7680b8aad4fa443b38ee59ab560c767f2f8f /source/convert.cc | |
| parent | 03510421970e0029bf744feac5c128c2ee22c90c (diff) | |
| download | libyuv-54ce8f23d6d82d671ec05229812842548faf14ef.tar.gz | |
Add MSA optimized ARGB/ABGR/BGRA/RGBA To Y/UV row functions
R=fbarchard@google.com
BUG=libyuv:634
Performance Gain (vs C auto-vectorized)
ARGBToYJRow_MSA - ~3.2x
ARGBToYJRow_Any_MSA - ~2.7x
BGRAToYRow_MSA - ~3.2x
BGRAToYRow_Any_MSA - ~2.7x
ABGRToYRow_MSA - ~3.2x
ABGRToYRow_Any_MSA - ~2.6x
RGBAToYRow_MSA - ~3.1x
RGBAToYRow_Any_MSA - ~2.7x
ARGBToUVJRow_MSA - ~5.5x
ARGBToUVJRow_Any_MSA - ~4.5x
BGRAToUVRow_MSA - ~2.1x
BGRAToUVRow_Any_MSA - ~2.0x
ABGRToUVRow_MSA - ~2.1x
ABGRToUVRow_Any_MSA - ~1.9x
RGBAToUVRow_MSA - ~2.2x
RGBAToUVRow_Any_MSA - ~1.9x
Performance Gain (vs C non-vectorized)
ARGBToYJRow_MSA - ~10.9x
ARGBToYJRow_Any_MSA - ~9.2x
BGRAToYRow_MSA - ~10.9x
BGRAToYRow_Any_MSA - ~9.3x
ABGRToYRow_MSA - ~11.0x
ABGRToYRow_Any_MSA - ~9.3x
RGBAToYRow_MSA - ~10.9x
RGBAToYRow_Any_MSA - ~9.1x
ARGBToUVJRow_MSA - ~12.4x
ARGBToUVJRow_Any_MSA - ~10.5x
BGRAToUVRow_MSA - ~4.7x
BGRAToUVRow_Any_MSA - ~4.4x
ABGRToUVRow_MSA - ~4.7x
ABGRToUVRow_Any_MSA - ~4.5x
RGBAToUVRow_MSA - ~4.8x
RGBAToUVRow_Any_MSA - ~4.4x
Review-Url: https://codereview.chromium.org/2641153003 .
Diffstat (limited to 'source/convert.cc')
| -rw-r--r-- | source/convert.cc | 48 |
1 files changed, 48 insertions, 0 deletions
diff --git a/source/convert.cc b/source/convert.cc index d5d12acb..f79acaca 100644 --- a/source/convert.cc +++ b/source/convert.cc @@ -696,6 +696,22 @@ int BGRAToI420(const uint8* src_bgra, } } #endif +#if defined(HAS_BGRATOYROW_MSA) + if (TestCpuFlag(kCpuHasMSA)) { + BGRAToYRow = BGRAToYRow_Any_MSA; + if (IS_ALIGNED(width, 16)) { + BGRAToYRow = BGRAToYRow_MSA; + } + } +#endif +#if defined(HAS_BGRATOUVROW_MSA) + if (TestCpuFlag(kCpuHasMSA)) { + BGRAToUVRow = BGRAToUVRow_Any_MSA; + if (IS_ALIGNED(width, 16)) { + BGRAToUVRow = BGRAToUVRow_MSA; + } + } +#endif for (y = 0; y < height - 1; y += 2) { BGRAToUVRow(src_bgra, src_stride_bgra, dst_u, dst_v, width); @@ -781,6 +797,22 @@ int ABGRToI420(const uint8* src_abgr, } } #endif +#if defined(HAS_ABGRTOYROW_MSA) + if (TestCpuFlag(kCpuHasMSA)) { + ABGRToYRow = ABGRToYRow_Any_MSA; + if (IS_ALIGNED(width, 16)) { + ABGRToYRow = ABGRToYRow_MSA; + } + } +#endif +#if defined(HAS_ABGRTOUVROW_MSA) + if (TestCpuFlag(kCpuHasMSA)) { + ABGRToUVRow = ABGRToUVRow_Any_MSA; + if (IS_ALIGNED(width, 16)) { + ABGRToUVRow = ABGRToUVRow_MSA; + } + } +#endif for (y = 0; y < height - 1; y += 2) { ABGRToUVRow(src_abgr, src_stride_abgr, dst_u, dst_v, width); @@ -866,6 +898,22 @@ int RGBAToI420(const uint8* src_rgba, } } #endif +#if defined(HAS_RGBATOYROW_MSA) + if (TestCpuFlag(kCpuHasMSA)) { + RGBAToYRow = RGBAToYRow_Any_MSA; + if (IS_ALIGNED(width, 16)) { + RGBAToYRow = RGBAToYRow_MSA; + } + } +#endif +#if defined(HAS_RGBATOUVROW_MSA) + if (TestCpuFlag(kCpuHasMSA)) { + RGBAToUVRow = RGBAToUVRow_Any_MSA; + if (IS_ALIGNED(width, 16)) { + RGBAToUVRow = RGBAToUVRow_MSA; + } + } +#endif for (y = 0; y < height - 1; y += 2) { RGBAToUVRow(src_rgba, src_stride_rgba, dst_u, dst_v, width); |