diff options
| author | Clément Bœsch <ubitux@gmail.com> | 2012-12-07 00:39:31 +0100 |
|---|---|---|
| committer | Anton Khirnov <anton@khirnov.net> | 2013-03-28 07:59:18 +0100 |
| commit | 38a2f88d39e50c573193b6d3f14df58b6c4e3a7b (patch) | |
| tree | 182ad061ef3b33ff84afb57646922489b192a018 | |
| parent | 2d66fc543b01995d6146fc132a778d3e722ca665 (diff) | |
| download | ffmpeg-38a2f88d39e50c573193b6d3f14df58b6c4e3a7b.tar.gz | |
lavfi/gradfun: fix dithering in MMX code.
Current dithering only uses the first 4 instead of the whole 8 random values.
Signed-off-by: Anton Khirnov <anton@khirnov.net>
| -rw-r--r-- | libavfilter/x86/vf_gradfun.c | 31 |
1 files changed, 29 insertions, 2 deletions
diff --git a/libavfilter/x86/vf_gradfun.c b/libavfilter/x86/vf_gradfun.c index 75c117a9a2..a9e069a5bf 100644 --- a/libavfilter/x86/vf_gradfun.c +++ b/libavfilter/x86/vf_gradfun.c @@ -46,7 +46,9 @@ static void gradfun_filter_line_mmxext(uint8_t *dst, uint8_t *src, uint16_t *dc, "pxor %%mm7, %%mm7 \n" "pshufw $0, %%mm5, %%mm5 \n" "movq %6, %%mm6 \n" - "movq %5, %%mm4 \n" + "movq (%5), %%mm3 \n" + "movq 8(%5), %%mm4 \n" + "1: \n" "movd (%2,%0), %%mm0 \n" "movd (%3,%0), %%mm1 \n" @@ -61,6 +63,29 @@ static void gradfun_filter_line_mmxext(uint8_t *dst, uint8_t *src, uint16_t *dc, "psubw %%mm6, %%mm2 \n" "pminsw %%mm7, %%mm2 \n" // m = -max(0, 127-m) "pmullw %%mm2, %%mm2 \n" + "paddw %%mm3, %%mm0 \n" // pix += dither + "psllw $2, %%mm1 \n" // m = m*m*delta >> 14 + "pmulhw %%mm2, %%mm1 \n" + "paddw %%mm1, %%mm0 \n" // pix += m + "psraw $7, %%mm0 \n" + "packuswb %%mm0, %%mm0 \n" + "movd %%mm0, (%1,%0) \n" // dst = clip(pix>>7) + "add $4, %0 \n" + "jnl 2f \n" + + "movd (%2,%0), %%mm0 \n" + "movd (%3,%0), %%mm1 \n" + "punpcklbw %%mm7, %%mm0 \n" + "punpcklwd %%mm1, %%mm1 \n" + "psllw $7, %%mm0 \n" + "pxor %%mm2, %%mm2 \n" + "psubw %%mm0, %%mm1 \n" // delta = dc - pix + "psubw %%mm1, %%mm2 \n" + "pmaxsw %%mm1, %%mm2 \n" + "pmulhuw %%mm5, %%mm2 \n" // m = abs(delta) * thresh >> 16 + "psubw %%mm6, %%mm2 \n" + "pminsw %%mm7, %%mm2 \n" // m = -max(0, 127-m) + "pmullw %%mm2, %%mm2 \n" "paddw %%mm4, %%mm0 \n" // pix += dither "psllw $2, %%mm1 \n" // m = m*m*delta >> 14 "pmulhw %%mm2, %%mm1 \n" @@ -70,10 +95,12 @@ static void gradfun_filter_line_mmxext(uint8_t *dst, uint8_t *src, uint16_t *dc, "movd %%mm0, (%1,%0) \n" // dst = clip(pix>>7) "add $4, %0 \n" "jl 1b \n" + + "2: \n" "emms \n" :"+r"(x) :"r"(dst+width), "r"(src+width), "r"(dc+width/2), - "rm"(thresh), "m"(*dithers), "m"(*pw_7f) + "rm"(thresh), "r"(dithers), "m"(*pw_7f) :"memory" ); } |