On Sun, Aug 31, 2014 at 06:28:24PM +0200, Reimar Döffinger wrote: > On Sun, Aug 31, 2014 at 05:44:09PM +0200, Michael Niedermayer wrote: > > On Sun, Aug 31, 2014 at 04:48:41PM +0200, Reimar Döffinger wrote: > > > Some additional optimizations in following patch. > > > > what effect on speed do these patches have ? > > With high error bars, with vsynth1 file, median of 3 runs, only around > changed code: > h261: > 4447 decicycles in a, 65532 runs, 4 skips > to > 3894 decicycles in a, 65534 runs, 2 skips > > h263: > 5072 decicycles in a, 65532 runs, 4 skips > to > 3502 decicycles in a, 65533 runs, 3 skips > > h263p: > 5257 decicycles in a, 65533 runs, 3 skips > to > 7880 decicycles in a, 65532 runs, 4 skips > > flv: > 5291 decicycles in a, 65532 runs, 4 skips > to > 3609 decicycles in a, 65531 runs, 5 skips > > Did not test the pure RL_VLC without optimizations. > I guess the H263P one really doesn't like the close and reopen (?)
Nope, just me sucking at testing (copy-pasted wrong line) Should read: h263p: 11460 decicycles in a, 65533 runs, 3 skips to 7880 decicycles in a, 65532 runs, 4 skips _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel