diff options
| author | Clément Bœsch <u@pkh.me> | 2014-01-14 08:09:48 +0100 |
|---|---|---|
| committer | Clément Bœsch <clement@stupeflix.com> | 2014-01-15 15:54:03 +0100 |
| commit | 8b4190da9382434758e390370b1752583bf4ce3a (patch) | |
| tree | da0db2f11156c6fabd71ea752f7203a055d78bf2 /libavcodec/x86/vp9itxfm.asm | |
| parent | 53e6977c07a8720cf4f785ef23686bf34b5cec57 (diff) | |
| download | ffmpeg-8b4190da9382434758e390370b1752583bf4ce3a.tar.gz | |
vp9/x86: add AVX for itxfm and lpf.
4412 decicycles in ff_vp9_loop_filter_h_16_16_ssse3, 4193462 runs, 842 skips
3600 decicycles in ff_vp9_loop_filter_h_16_16_avx, 4193621 runs, 683 skips
3010 decicycles in ff_vp9_loop_filter_v_16_16_ssse3, 4193528 runs, 776 skips
2678 decicycles in ff_vp9_loop_filter_v_16_16_avx, 4193742 runs, 562 skips
23025 decicycles in ff_vp9_idct_idct_32x32_add_ssse3, 2096871 runs, 281 skips
19943 decicycles in ff_vp9_idct_idct_32x32_add_avx, 2096815 runs, 337 skips
4675 decicycles in ff_vp9_idct_idct_16x16_add_ssse3, 4194018 runs, 286 skips
3980 decicycles in ff_vp9_idct_idct_16x16_add_avx, 4194022 runs, 282 skips
967 decicycles in ff_vp9_idct_idct_8x8_add_ssse3, 16776972 runs, 244 skips
887 decicycles in ff_vp9_idct_idct_8x8_add_avx, 16777002 runs, 214 skips
Diffstat (limited to 'libavcodec/x86/vp9itxfm.asm')
| -rw-r--r-- | libavcodec/x86/vp9itxfm.asm | 21 |
1 files changed, 18 insertions, 3 deletions
diff --git a/libavcodec/x86/vp9itxfm.asm b/libavcodec/x86/vp9itxfm.asm index fe04b81a26..33c0bc790b 100644 --- a/libavcodec/x86/vp9itxfm.asm +++ b/libavcodec/x86/vp9itxfm.asm @@ -289,7 +289,8 @@ cglobal vp9_idct_idct_4x4_add, 4,4,0, dst, stride, block, eob VP9_STORE_2X 10, 11, 6, 7, 4 %endmacro -INIT_XMM ssse3 +%macro VP9_IDCT_IDCT_8x8_ADD_XMM 1 +INIT_XMM %1 cglobal vp9_idct_idct_8x8_add, 4,4,13, dst, stride, block, eob mova m12, [pw_11585x2] ; often used @@ -376,6 +377,10 @@ cglobal vp9_idct_idct_8x8_add, 4,4,13, dst, stride, block, eob mova [blockq+112], m4 VP9_IDCT8_WRITEOUT RET +%endmacro + +VP9_IDCT_IDCT_8x8_ADD_XMM ssse3 +VP9_IDCT_IDCT_8x8_ADD_XMM avx ;--------------------------------------------------------------------------------------------- ; void vp9_idct_idct_16x16_add_<opt>(uint8_t *dst, ptrdiff_t stride, int16_t *block, int eob); @@ -655,7 +660,8 @@ cglobal vp9_idct_idct_8x8_add, 4,4,13, dst, stride, block, eob mova [dstq+%7], m%4 %endmacro -INIT_XMM ssse3 +%macro VP9_IDCT_IDCT_16x16_ADD_XMM 1 +INIT_XMM %1 cglobal vp9_idct_idct_16x16_add, 4, 5, 16, 512, dst, stride, block, eob ; 2x2=eob=3, 4x4=eob=10 cmp eobd, 38 @@ -724,6 +730,10 @@ cglobal vp9_idct_idct_16x16_add, 4, 5, 16, 512, dst, stride, block, eob ; use that to zero out block coefficients ZERO_BLOCK blockq, 32, 16, m0 RET +%endmacro + +VP9_IDCT_IDCT_16x16_ADD_XMM ssse3 +VP9_IDCT_IDCT_16x16_ADD_XMM avx ;--------------------------------------------------------------------------------------------- ; void vp9_idct_idct_32x32_add_<opt>(uint8_t *dst, ptrdiff_t stride, int16_t *block, int eob); @@ -1102,7 +1112,8 @@ cglobal vp9_idct_idct_16x16_add, 4, 5, 16, 512, dst, stride, block, eob %endif %endmacro -INIT_XMM ssse3 +%macro VP9_IDCT_IDCT_32x32_ADD_XMM 1 +INIT_XMM %1 cglobal vp9_idct_idct_32x32_add, 4, 8, 16, 2048, dst, stride, block, eob cmp eobd, 135 jg .idctfull @@ -1213,5 +1224,9 @@ cglobal vp9_idct_idct_32x32_add, 4, 8, 16, 2048, dst, stride, block, eob ; use that to zero out block coefficients ZERO_BLOCK blockq, 64, 32, m7 RET +%endmacro + +VP9_IDCT_IDCT_32x32_ADD_XMM ssse3 +VP9_IDCT_IDCT_32x32_ADD_XMM avx %endif ; x86-64 |