diff options
| author | Ronald S. Bultje <rsbultje@gmail.com> | 2014-01-20 17:11:48 -0500 |
|---|---|---|
| committer | Ronald S. Bultje <rsbultje@gmail.com> | 2014-01-24 19:25:25 -0500 |
| commit | baf47020cd2309e57f73883b703b5e6c36c7f4df (patch) | |
| tree | 8fb132286ae57f8b4258a35b8603dba88de001c9 /libavcodec/x86/vp9dsp_init.c | |
| parent | 2efdccac87cbcf8f9ef3cd12201fed0e0c93f02b (diff) | |
| download | ffmpeg-baf47020cd2309e57f73883b703b5e6c36c7f4df.tar.gz | |
vp9/x86: 8x8 iadst SIMD (ssse3/avx) variants.
Cycle measurements for intra itxfm_8x8_add on ped1080p.webm:
idct_idct: 133 -> 135 cycles (noise measurement)
idct_iadst: 900 -> 241 cycles
iadst_idct: 864 -> 215 cycles
iadst_iadst: 973 -> 310 cycles
Diffstat (limited to 'libavcodec/x86/vp9dsp_init.c')
| -rw-r--r-- | libavcodec/x86/vp9dsp_init.c | 10 |
1 files changed, 8 insertions, 2 deletions
diff --git a/libavcodec/x86/vp9dsp_init.c b/libavcodec/x86/vp9dsp_init.c index f6a213529c..a61b589b4e 100644 --- a/libavcodec/x86/vp9dsp_init.c +++ b/libavcodec/x86/vp9dsp_init.c @@ -167,8 +167,8 @@ itxfm_func(idct, iadst, size, opt); \ itxfm_func(iadst, iadst, size, opt) itxfm_func(idct, idct, 4, ssse3); -itxfm_func(idct, idct, 8, ssse3); -itxfm_func(idct, idct, 8, avx); +itxfm_funcs(8, ssse3); +itxfm_funcs(8, avx); itxfm_funcs(16, ssse3); itxfm_funcs(16, avx); itxfm_func(idct, idct, 32, ssse3); @@ -252,6 +252,9 @@ av_cold void ff_vp9dsp_init_x86(VP9DSPContext *dsp) dsp->itxfm_add[TX_4X4][DCT_DCT] = ff_vp9_idct_idct_4x4_add_ssse3; if (ARCH_X86_64) { dsp->itxfm_add[TX_8X8][DCT_DCT] = ff_vp9_idct_idct_8x8_add_ssse3; + dsp->itxfm_add[TX_8X8][ADST_DCT] = ff_vp9_idct_iadst_8x8_add_ssse3; + dsp->itxfm_add[TX_8X8][DCT_ADST] = ff_vp9_iadst_idct_8x8_add_ssse3; + dsp->itxfm_add[TX_8X8][ADST_ADST] = ff_vp9_iadst_iadst_8x8_add_ssse3; dsp->itxfm_add[TX_16X16][DCT_DCT] = ff_vp9_idct_idct_16x16_add_ssse3; dsp->itxfm_add[TX_16X16][ADST_DCT] = ff_vp9_idct_iadst_16x16_add_ssse3; dsp->itxfm_add[TX_16X16][DCT_ADST] = ff_vp9_iadst_idct_16x16_add_ssse3; @@ -268,6 +271,9 @@ av_cold void ff_vp9dsp_init_x86(VP9DSPContext *dsp) if (EXTERNAL_AVX(cpu_flags)) { if (ARCH_X86_64) { dsp->itxfm_add[TX_8X8][DCT_DCT] = ff_vp9_idct_idct_8x8_add_avx; + dsp->itxfm_add[TX_8X8][ADST_DCT] = ff_vp9_idct_iadst_8x8_add_avx; + dsp->itxfm_add[TX_8X8][DCT_ADST] = ff_vp9_iadst_idct_8x8_add_avx; + dsp->itxfm_add[TX_8X8][ADST_ADST] = ff_vp9_iadst_iadst_8x8_add_avx; dsp->itxfm_add[TX_16X16][DCT_DCT] = ff_vp9_idct_idct_16x16_add_avx; dsp->itxfm_add[TX_16X16][ADST_DCT] = ff_vp9_idct_iadst_16x16_add_avx; dsp->itxfm_add[TX_16X16][DCT_ADST] = ff_vp9_iadst_idct_16x16_add_avx; |