Mercurial > libavcodec.hg
diff x86/vp8dsp-init.c @ 12241:c7f6ddcc5c01 libavcodec
VP8: optimize DC-only chroma case in the same way as luma.
Add MMX idct_dc_add4uv function for this case.
~40% faster chroma idct.
| author | darkshikari |
|---|---|
| date | Fri, 23 Jul 2010 06:02:52 +0000 |
| parents | 1a7903913e9b |
| children | 435319d67bd8 |
line wrap: on
line diff
--- a/x86/vp8dsp-init.c Fri Jul 23 03:44:37 2010 +0000 +++ b/x86/vp8dsp-init.c Fri Jul 23 06:02:52 2010 +0000 @@ -220,8 +220,9 @@ 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_idct_dc_add4y_mmx(uint8_t *dst, DCTELEM block[4][16], int stride); +extern void ff_vp8_idct_dc_add4y_sse2(uint8_t *dst, DCTELEM block[4][16], int stride); +extern void ff_vp8_idct_dc_add4uv_mmx(uint8_t *dst, DCTELEM block[2][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); @@ -284,10 +285,11 @@ #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->vp8_idct_dc_add = ff_vp8_idct_dc_add_mmx; + c->vp8_idct_dc_add4y = ff_vp8_idct_dc_add4y_mmx; + c->vp8_idct_dc_add4uv = ff_vp8_idct_dc_add4uv_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] = @@ -354,7 +356,7 @@ } if (mm_flags & FF_MM_SSE2) { - c->vp8_idct_dc_add4 = ff_vp8_idct_dc_add4_sse2; + c->vp8_idct_dc_add4y = ff_vp8_idct_dc_add4y_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;