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/vp8dsp.c | |
| parent | ef38842f0bc97ce5b158f51f3e65aae4164fc6a5 (diff) | |
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/vp8dsp.c')
| -rw-r--r-- | libavcodec/vp8dsp.c | 26 |
1 files changed, 23 insertions, 3 deletions
diff --git a/libavcodec/vp8dsp.c b/libavcodec/vp8dsp.c index 3e6463d598..64b09d52ee 100644 --- a/libavcodec/vp8dsp.c +++ b/libavcodec/vp8dsp.c @@ -109,6 +109,25 @@ static void vp8_idct_dc_add_c(uint8_t *dst, DCTELEM block[16], int stride) } } +static void vp8_idct_dc_add4_c(uint8_t *dst, DCTELEM block[4][16], int stride) +{ + int i, j; + for (j = 0; j < 4; j++) { + uint8_t *pix = dst+j*4; + int dc = (block[j][0] + 4) >> 3; + uint8_t *cm = ff_cropTbl + MAX_NEG_CROP + dc; + block[j][0] = 0; + if (!dc) + continue; + for (i = 0; i < 4; i++) { + pix[0] = cm[pix[0]]; + pix[1] = cm[pix[1]]; + pix[2] = cm[pix[2]]; + pix[3] = cm[pix[3]]; + pix += stride; + } + } +} // because I like only having two parameters to pass functions... #define LOAD_PIXELS\ @@ -460,9 +479,10 @@ VP8_BILINEAR(4) av_cold void ff_vp8dsp_init(VP8DSPContext *dsp) { - dsp->vp8_luma_dc_wht = vp8_luma_dc_wht_c; - dsp->vp8_idct_add = vp8_idct_add_c; - dsp->vp8_idct_dc_add = vp8_idct_dc_add_c; + dsp->vp8_luma_dc_wht = vp8_luma_dc_wht_c; + dsp->vp8_idct_add = vp8_idct_add_c; + dsp->vp8_idct_dc_add = vp8_idct_dc_add_c; + dsp->vp8_idct_dc_add4 = vp8_idct_dc_add4_c; dsp->vp8_v_loop_filter16y = vp8_v_loop_filter16_c; dsp->vp8_h_loop_filter16y = vp8_h_loop_filter16_c; |