[FFmpeg-devel] (new) intermediate codec

[Michael Niedermayer]

On Thu, Mar 10, 2011 at 10:47:03PM +0100, Lars T?uber wrote:
> On Thu, 10 Mar 2011 22:28:40 +0100 Michael Niedermayer <michaelni at gmx.at> wrote:
> > On Thu, Mar 10, 2011 at 09:23:35PM +0100, Lars T?uber wrote:
> > > On Thu, 10 Mar 2011 21:02:38 +0100 Michael Niedermayer <michaelni at gmx.at> wrote:
> > > > On Thu, Mar 10, 2011 at 08:58:00PM +0100, Lars T?uber wrote:
> > > > > Hi Michael,
> > > > > 
> > > > > On Thu, 10 Mar 2011 10:03:43 +0100 Michael Niedermayer <michaelni at gmx.at> wrote:
> > > > > > Dont forget testing multithreaded ffv1.2
> > > > > > 
> > > > > 
> > > > > is this irony or do you really work on a multithreading capable version of ffv1?
> > > > 
> > > > set version=2 in ffv1.c
> > > > 
> > > 
> > > I just had a short look.
> > 
> > I am just the author
> > 
> 
> I know.
> 
> > 
> > > 
> > > version=2 seems to be disabled on purpose somehow:
> > > line 1010:
> > >     av_assert0(s->version>=2);
> > > 
> > 
> > > And is this really scaling on a modern cpu? This reads like it supports max 4 threads:
> > >     if(s->version>1){
> > >         s->num_h_slices=2;
> > >         s->num_v_slices=2;
> > >         write_extra_header(s);
> > >     }
> > 
> > You can put larger numbers in there
> 
> This would mean the scalability is a compile time option. You had to assure the width and/or hight is dividable by that number.
> I'd prefer frame based multithreading.
> 
> > version 2 is disabled because its work in progress code, once finished these
> > would be accessable through the command line of course
> 
> I just thought that. But the progress is slow right now.
> Decoding speed is really important for an intermediate codec. Could there be some improvement with ffv1? It is not usable with HD video on my 2,5GHz 2 core machine. I can use HuffYUV with FullHD just fine on this machine. So I thought this could be a good starting point for a codec with the requested features.

Iam quite busy ATM to improve it myself, but i surely could integrate patches :)

[...]
--
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

it is not once nor twice but times without number that the same ideas make
their appearance in the world. -- Aristotle
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20110310/dd0ef253/attachment.pgp>