Hi, On Sat, Apr 16, 2011 at 3:27 PM, Kostya <[email protected]> wrote: > On Sat, Apr 16, 2011 at 03:19:43PM -0400, Ronald S. Bultje wrote: >> On Sat, Apr 16, 2011 at 3:08 PM, Ronald S. Bultje <[email protected]> wrote: >> > On Sat, Apr 16, 2011 at 3:04 PM, Kostya <[email protected]> wrote: >> >> in general looks ok but won't it cause noticeable performance drop? >> >> If yes than maybe it's better to use single flag and have two codepaths >> >> with >> >> constants. >> > >> > Maybe. I figured that it might, and wrote the patch halfway adding the >> > two codepaths depending on the res_fasttx flag, but it makes it look >> > pretty clumsy, so I decided to do it this way. Should be OK. If we >> > ever want to do everything and the world to optimize wmv3/vc1, we can >> > expand this into constant versions under if/else. >> >> before: user 0m47.125s >> after: user 0m47.088s >> >> So a negligible speed difference. The difference you're seeing is most >> likely noise, because it should get slower, not faster... > > ok then (don't forget to remove stray whitespace before applying though)
Pushed. Ronald _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
