[XviD-devel] On the road to 1.0

[Michael Militzer]

Hi,

Quoting Edouard Gomez <ed.gomez at free.fr>:

> Christoph Lampert (chl at math.uni-bonn.de) wrote:
> > 
> > Hi,
> > 
> > to all of you, what is still missing for a release of dev-api-4 as 
> > XVID 1.0 ? 
> > 
> > 1) Full GMC support for ASP compatibility  (finished, not commited)
> >       * will be, after dev-api-4 encoder loop is stable.
> > 
> > 2) Windows GUI - simple presets/defaults
> >      * two-pass? 
> > 
> > 3) Native API - simple way to set flags 
> >      * two-pass?
> > 
> > Anything else that is _really_ needed? 
> 
> 4) Yep,  the  quaterpel  code  from  the  Isibaar  branch  needs  to  be
>    integrated.  What else  is in  this  mysterious Isiibar  branch ?  oh
>    yeah... the fast bframes.

-qpel
-fast b-frames
-cartoon mode
-postprocessing

Postprocessing still needs to be mmx'ed. Fast b-frames are not as fast as 
they could be, because the dominant factor for b-frames is ME. We should try 
to reduce the number of ME passes needed for b-frames (as I once proposed:
no special ME-analysis, but do a normal search and use the results to save 
the forward search-step (for b-frames and the whole ME for p-frames)...

> 4bis) And the  actual code is buggy,  some frames are  badly compensated. I
>    did not tracked deeply the bug  so i cannot tell if it occurs because
>    of wrongs  vectors in ME  or because of  another parasiting bug  in a
>    later stage. 

Hm, it shouldn't be buggy. My local tree worked without problems. But I 
applied my changes against a fresh checkout and maybe I forgot to commit
something...

> 5) 2pass does not work  for me(tm). Pete, does it work for  you ? If not
>    then it is really important to make it work. 

5 a) credits handling in 2pass encoding destroys file size accuracy. I'd vote 
to remove the credits feature, it has nothing to do with a video codec but
rather belongs to a support tool.

5 b) 2pass code is broken when the "packed bitstream" option is checked. 
This needs to be fixed, because I consider "packed bitstream" a really 
useful feature...
 
> 6) the  single pass  plugin needs  to integrate  the fixed  quant plugin
>    functionnality. See my last mail:
>    http://list.xvid.org/pipermail/xvid-devel/2003-May/002906.html
> 
> 7) Legal headers must make their come back. Current cvs code (except the
>    release-0_9_[01][-fixes]  tags/branches)  is   a  legal  mess,  wrong
>    license headers, no copyrights.
> 
> 8) I did an effort in 0.9.x to  document the API with a not so intrusive
>    method: doxygen. Do  we continue this way or do we  decide to write a
>    real documentation ? (texinfo, docbook xml ?)
> 
> All  steps before  8 should  be done  before the  dev-api-4  -> cvs_head
> merge imo.

yes.

bye,
Michael