On Fri, May 22, 2015 at 11:59:54PM +0200, Michael Niedermayer wrote: > this increases mpeg4 and msmpeg4v3 decoding speed by about 0.3-0.5% overall > > This reverts the merges of 848e86f74d3e6e87fa592ee8ba8c184cc5fd9a42 and > 9c1db92ad372d4cd69e0490e691c56e4097cb193 > > Speed tested with decoding matrixbench_mpg4.avi and matrixbench_msmpeg4v3.avi > Test files where generated as in > ./ffmpeg -i matrixbench_mpeg2.mpg -vb 1000k -vcodec mpeg4 matrixbench_mpg4.avi > > benchmarked with: > time ./ffmpeg -an -i matrixbench_mpg4.avi -an -f null - > > benchmarks: (each is best time of 7) > the values fluctuate but its consistently faster when reverted > > new 0m1.910s > reverted 0m1.901s > > new 1.907 > reverted 1.904 > > new 1.908 > reverted 1.907 > > new 0.461 > reverted 0.459 > > new 0.461 > reverted 0.459 > > new 0.461 > reverted 0.457 > > new 0.458 > reverted 0.456 >
Can't you just make a local (to the function) copy of the variable in the functions where it matters and gain back that negligible speedup? [...] -- Clément B.
pgptN07vmR095.pgp
Description: PGP signature
_______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel