On date Tuesday 2014-12-23 22:22:41 +0100, Michael Niedermayer encoded: > On Tue, Dec 23, 2014 at 10:51:09PM +0530, arwa arif wrote: > [...] > > > Test with a fixed qp value for comparison, as the mp=fspp is broken > > > with regards to qp passing. Also it should generate the same > > > output as with -cpuflags 0, but only in case the original filter > > > did. Report in any case. > > > > > > > mp=fspp and fspp filter are giving same output. But, fspp with -cpuflags 0 > > is giving different output (visually, it > > looks same, but its not bitexact). > > mp-fspp matches fspp with cpuflags 0 but mp-fspp ignores -cpuflags > so to compare HAVE_MMX_INLINE must be forced to 0 > > MMX mp-fspp == MMX fspp > > != > > C mp-fspp == C fspp > > > patch LGTM
With latest patch: ffmpeg -cpuflags 0 -i matrixbench_mpeg2-lq.mpg -vf fspp=5:20 -t 2 -f md5 - [...] MD5=fc74069dbf39742227bcb125ad657b5c 0kB time=00:00:02.00 bitrate= 0.0kbits/s ffmpeg -i matrixbench_mpeg2-lq.mpg -vf fspp=5:20 -t 2 -f md5 - [...] MD5=b65efd115976ecfdb942bd7cb1f0d967 0kB time=00:00:02.00 bitrate= 0.0kbits/s with #define HAVE_MMX_INLINE 0 set in the sourcecode: ffmpeg -cpuflags 0 -i matrixbench_mpeg2-lq.mpg -vf mp=fspp=5:20 -t 2 -f md5 - MD5=fc74069dbf39742227bcb125ad657b5c 0kB time=00:00:02.00 bitrate= 0.0kbits/s ffmpeg -i matrixbench_mpeg2-lq.mpg -vf mp=fspp=5:20 -t 2 -f md5 - [...] MD5=b65efd115976ecfdb942bd7cb1f0d967 0kB time=00:00:02.00 bitrate= 0.0kbits/s It would be interesting to understand why they MMX and C version differ, but this would be beside the scope of the port. I'm going to apply the patch soon, thanks. -- FFmpeg = Fast & Freak Murdering Peaceless Eager Gnome _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel