[FFmpeg-devel] [PATCH] increase MAX_AUDIO_FRAMESIZE

Kostya kostya.shishkov
Thu Sep 3 05:47:28 CEST 2009


On Wed, Sep 02, 2009 at 09:16:26PM +0200, Sascha Sommer wrote:
> Hi,
> 
> On Mittwoch, 2. September 2009, Michael Niedermayer wrote:
> > On Wed, Sep 02, 2009 at 08:03:02PM +0300, Kostya wrote:
> > > On Wed, Sep 02, 2009 at 07:00:32PM +0200, Sascha Sommer wrote:
> > > > Hi,
> > > >
> > > > attached patch increases the maximum audio framesizes. This makes it
> > > > possible to play a few more wmapro files.
> > >
> > > And WavPack too :)
> > >
> > > > Regards
> > > >
> > > > Sascha
> > >
> > > I wonder how long we will increase that buffer before someone proposes
> > > better buffer allocation system.
> >
> > i wonder if the decoders could maybe output smaller chunks?
> > its better memory wise, better cache wise and also improves
> > AV sync buffering requirements
> >
> 
> I wonder if it would help if the decode function could output smaller chunks 
> and set some flag so that it gets called again with the same AVPacket?
> What I want to avoid is to copy either the decoded or the nondecoded data into 
> some large temporary buffer. For wmapro I think it would be the best if 
> decoding is stopped after one "frame". This would require 8 * 4096 
> sizeof(float) bytes for an 8 channel stream as output buffer. At this point 
> only a part of the input buffer is consumed. Therefore it would be nice to 
> have the same AVPacket again without reinitializing get_bits etc. 
> What are the requirements for WavPack?

WavPack just may have large block of samples and a variety of bitdepths,
but you can stop decoding at any time and get samples. I'll think on how
to make it output lesser chunks.
 
> Regards
> 
> Sascha



More information about the ffmpeg-devel mailing list