diff options
| author | Jason Garrett-Glaser <darkshikari@gmail.com> | 2010-07-21 20:51:01 +0000 |
|---|---|---|
| committer | Jason Garrett-Glaser <darkshikari@gmail.com> | 2010-07-21 20:51:01 +0000 |
| commit | b8b231b5dc104192c0b766798a04cdf6e748472f (patch) | |
| tree | 1631715ed493ab17b3f687251a0930644dac2be3 /libavcodec/x86/vp8dsp-init.c | |
| parent | a4e6fa86ce7b3e4c7795fb725fa1de1b8dff15b4 (diff) | |
| download | ffmpeg-b8b231b5dc104192c0b766798a04cdf6e748472f.tar.gz | |
Make mmx VP8 WHT faster
Avoid pextrw, since it's slow on many older CPUs.
Now it doesn't require mmxext either.
Originally committed as revision 24397 to svn://svn.ffmpeg.org/ffmpeg/trunk
Diffstat (limited to 'libavcodec/x86/vp8dsp-init.c')
| -rw-r--r-- | libavcodec/x86/vp8dsp-init.c | 4 |
1 files changed, 2 insertions, 2 deletions
diff --git a/libavcodec/x86/vp8dsp-init.c b/libavcodec/x86/vp8dsp-init.c index 00f4bfbd15..a7b1ce0de7 100644 --- a/libavcodec/x86/vp8dsp-init.c +++ b/libavcodec/x86/vp8dsp-init.c @@ -220,7 +220,7 @@ 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_luma_dc_wht_mmxext(DCTELEM block[4][4][16], DCTELEM dc[16]); +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_v_loop_filter_simple_mmx (uint8_t *dst, int stride, int flim); @@ -315,6 +315,7 @@ av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c) if (mm_flags & FF_MM_MMX) { c->vp8_idct_dc_add = ff_vp8_idct_dc_add_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] = c->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_mmx; c->put_vp8_epel_pixels_tab[1][0][0] = @@ -337,7 +338,6 @@ av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c) /* note that 4-tap width=16 functions are missing because w=16 * is only used for luma, and luma is always a copy or sixtap. */ if (mm_flags & FF_MM_MMX2) { - c->vp8_luma_dc_wht = ff_vp8_luma_dc_wht_mmxext; VP8_LUMA_MC_FUNC(0, 16, mmxext); VP8_MC_FUNC(1, 8, mmxext); VP8_MC_FUNC(2, 4, mmxext); |