On Fri, Aug 06, 2010 at 05:56:25PM +0200, Reimar Döffinger wrote:
> On Fri, Aug 06, 2010 at 05:07:14PM +0200, Diego Biurrun wrote:
> > On Fri, Aug 06, 2010 at 10:16:49AM -0400, Reinhard Tartler wrote:
> > > 
> > > In case this works, Diego, Reimar, do you think it's worth to ship a
> > > different codecs.conf on arm-ish (arm, armel and armhf) platforms that
> > > prefer tremor over ffvorbis? Or can this preference perhaps be
> > > influenced by a configure switch?
> > 
> > I have thought about this before, but I have no good idea how to solve
> > this.  A different codecs.conf will work, but it's ugly.  Maybe we
> > could add a fixedpoint flag to codecs.conf entries and prefer these
> > codecs on systems without FPU?
> > 
> >  Reimar?
> I think the approach that was discussed for FFmpeg was to only
> compile the faster variant.
> I has quite a few issues, but in principle only compiling in
> tremor support would select it.
> It would also work the other way round (and probably better),
> make tremor the default but only compile it in on FPU-less
> systems.
> In the case of MPlayer there may be some issues with that,
> e.g. the native ogg demuxer behaviour might change.
> Of course the best solution would have been if tremor
> hadn't been developed as a completely different library but
> instead was just a different configuration of libvorbis...

I disagree slightly here since the issue does not only apply to
Vorbis/Tremor.  For MP3 we have a similar situation: We default
to mp3lib, but ffmp3 is fixedpoint and thus faster on systems
without FPU.  So a slightly more general framework on our side
might be adequate.


pkg-multimedia-maintainers mailing list

Reply via email to