On Thu, Apr 18, 2013 at 07:41:17PM +0300, Martin Storsjö wrote: > On Thu, 18 Apr 2013, Diego Biurrun wrote: > > >So, does anybody still care about the sh4 optimizations that we have? > >Did anybody run benchmarks to confirm they actually provide a speedup? > >Do we have a volunteer to set up a FATE instance on the hardware? > > > >Otherwise I think we should seriously consider just throwing the > >code away. It's ugly and crufty to say the least and it causes > >maintenance overhead. Whether it is worth the trouble is not > >clear at all. > > Most of these optimizations are just tweaks on plain C level, so I'm > a bit doubtful to how useful they are at this point, given the > compiler advances during the last 10 years. The few assembly bits > might be a bit more valuable.
Yes, I should have been more clear: There is an assembler IDCT and some bit shifting routines in libavutil. That stuff is probably worth keeping, but the compiler trickery is dubious. Diego _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
