[FFmpeg-devel] Proposed patch

Roman Shaposhnick rvs
Sat Aug 4 03:55:15 CEST 2007

On Sat, 2007-08-04 at 02:40 +0200, Michael Niedermayer wrote:
> Hi
> On Thu, Aug 02, 2007 at 08:22:02PM -0700, Roman Shaposhnik wrote:
> > Guys,
> > 
> > two questions: 
> >    1. Does an attached patch look ok?
> this changes one of the most speed critical things in lavc so this needs
> extensive benchmarks, x86+ppc mpeg4 decoding at minimum

  I can cover x86,x64 and SPARC. Would that be enough? Also, I can
actually look at assembly generated before and after on these arcs.
I actually expect it to be identical. 

  Anyway, if you consider ppc to be a critical one to benchmark on,
I would need somebody to do it for me.  

> (more non x86 archs would be very welcome too)
> also changing HAVE_FAST_UNALIGNED for negatively affected archs
> might prevent the speedloss
> >    2. I would like to add AV_ prefix to (LD|ST)(16|32|64) macros from 
> >       intreadwrite.h. The question is would it be enough or should
> >       we make them more similar to existing AV_(R|W)(B|L)(16|32|64)
> >       macros by going:
> >                LD(16|32|64) -> AV_R(16|32|64)
> >                ST(16|32|64) -> AV_W(16|32|64)
> i agree with the AV_ prefix and i dont really have any strong oppinion
> on the R/W vs. LD/ST stuff though i would prefer RN/WN for native endian
> over R/W

  Thanks. If there are no other opinions I'm going to settle on:
          LD(16|32|64) -> AV_RN(16|32|64)
          ST(16|32|64) -> AV_WN(16|32|64)


More information about the ffmpeg-devel mailing list