diff options
| author | Jason Garrett-Glaser <darkshikari@gmail.com> | 2010-07-23 02:58:27 +0000 |
|---|---|---|
| committer | Jason Garrett-Glaser <darkshikari@gmail.com> | 2010-07-23 02:58:27 +0000 |
| commit | 8a467b2d44d20c1a0b731dce9edeff772732a558 (patch) | |
| tree | 259b69473524a9db002d79b90009817f84af4113 /libavcodec/x86/vp8dsp-init.c | |
| parent | ef38842f0bc97ce5b158f51f3e65aae4164fc6a5 (diff) | |
| download | ffmpeg-8a467b2d44d20c1a0b731dce9edeff772732a558.tar.gz | |
VP8: 30% faster idct_mb
Take shortcuts based on statistically common situations.
Add 4-at-a-time idct_dc function (mmx and sse2) since rows of 4 DC-only DCT
blocks are common.
TODO: tie this more directly into the MB mode, since the DC-level transform is
only used for non-splitmv blocks?
Originally committed as revision 24452 to svn://svn.ffmpeg.org/ffmpeg/trunk
Diffstat (limited to 'libavcodec/x86/vp8dsp-init.c')
| -rw-r--r-- | libavcodec/x86/vp8dsp-init.c | 5 |
1 files changed, 5 insertions, 0 deletions
diff --git a/libavcodec/x86/vp8dsp-init.c b/libavcodec/x86/vp8dsp-init.c index 6cf1704594..5da70824fc 100644 --- a/libavcodec/x86/vp8dsp-init.c +++ b/libavcodec/x86/vp8dsp-init.c @@ -220,6 +220,8 @@ HVBILIN(ssse3, 8, 16, 16) extern void ff_vp8_idct_dc_add_mmx(uint8_t *dst, DCTELEM block[16], int stride); extern void ff_vp8_idct_dc_add_sse4(uint8_t *dst, DCTELEM block[16], int stride); +extern void ff_vp8_idct_dc_add4_mmx(uint8_t *dst, DCTELEM block[4][16], int stride); +extern void ff_vp8_idct_dc_add4_sse2(uint8_t *dst, DCTELEM block[4][16], int stride); extern void ff_vp8_luma_dc_wht_mmx(DCTELEM block[4][4][16], DCTELEM dc[16]); extern void ff_vp8_idct_add_mmx(uint8_t *dst, DCTELEM block[16], int stride); extern void ff_vp8_idct_add_sse(uint8_t *dst, DCTELEM block[16], int stride); @@ -283,6 +285,7 @@ av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c) #if HAVE_YASM if (mm_flags & FF_MM_MMX) { c->vp8_idct_dc_add = ff_vp8_idct_dc_add_mmx; + c->vp8_idct_dc_add4 = ff_vp8_idct_dc_add4_mmx; c->vp8_idct_add = ff_vp8_idct_add_mmx; c->vp8_luma_dc_wht = ff_vp8_luma_dc_wht_mmx; c->put_vp8_epel_pixels_tab[0][0][0] = @@ -351,6 +354,8 @@ av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c) } if (mm_flags & FF_MM_SSE2) { + c->vp8_idct_dc_add4 = ff_vp8_idct_dc_add4_sse2; + c->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16y_inner_sse2; c->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_sse2; |