On Sun, Aug 23, 2015 at 11:58:23AM -0400, Ganesh Ajjanagadde wrote: [...] > diff --git a/libavfilter/vf_hqx.c b/libavfilter/vf_hqx.c > index fa15d9c..0178793 100644 > --- a/libavfilter/vf_hqx.c > +++ b/libavfilter/vf_hqx.c > @@ -65,9 +65,9 @@ static av_always_inline int yuv_diff(uint32_t yuv1, > uint32_t yuv2) > #define YMASK 0xff0000 > #define UMASK 0x00ff00 > #define VMASK 0x0000ff > - return abs((yuv1 & YMASK) - (yuv2 & YMASK)) > (48 << 16) || > - abs((yuv1 & UMASK) - (yuv2 & UMASK)) > ( 7 << 8) || > - abs((yuv1 & VMASK) - (yuv2 & VMASK)) > ( 6 << 0); > + return FFUABSDIFF(yuv1 & YMASK, yuv2 & YMASK) > (48 << 16) || > + FFUABSDIFF(yuv1 & UMASK, yuv2 & UMASK) > ( 7 << 8) || > + FFUABSDIFF(yuv1 & VMASK, yuv2 & VMASK) > ( 6 << 0);
This is one of the bottleneck function of the filter. How does it affect speed? Can you compare the generated ASM? Same question for xbr. [...] -- Clément B.
pgpSm9FOXhodi.pgp
Description: PGP signature
_______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
