[x264-devel] NAL slice support patch added?

[Mojtaba Hosseini]

On Mon, 2007-10-22 at 22:53 -0600, Loren Merritt wrote:
> On Mon, 22 Oct 2007, Mojtaba Hosseini wrote:
> 
> > 2) The current patch does not do anything specific for rate control. 
> > Is this ok? That is, from the code, I do not see how this patch 
> > could affect rate control but please advise if otherwise.
> 
> Now that slices are divorced from threads, ratecontrol shouldn't care 
> about slices.
> 
> > 3) As per suggestion given on irc, number of NALs is allocated
> > dynamically (because number of NALs in frame can potentially be
> > very large). The default allocation is 20 and it grows by steps
> > of 20, every time there is a need to grow. Is this a reasonable
> > approach? Is 20 a good step size?
> 

Hello,
  I have changed the patch to include suggestions given by Loren and
Kurosu (thank you for looking over the code) but using the regression
test suggested by Loren (ie encode with x264, decode with JM and compare
 fdec.yuv to output of JM), I have discovered a problem: 
  Everything is fine (ie fdec.yuv and output of JM match perfectly) 
if the slice break happens at beginning of mb row (row = one horizontal
row of macroblocks starting at left edge and ending at right edge).
However there are differences when the slice break happens in the middle
of a mb row.  Can anyone guide me in the general direction of where I
should be looking to fix this? Loop filter? intra prediction for a mb
 from a previous slice (i assume not allowed)? 
  Any help is much appreciated. 

Mojtaba