[x264-devel] Re: why is deblocking one row placed in the encoding loop?

[Jindrich Makovicka]

On 5/30/07, Alex Izvorski <aizvorski at gmail.com> wrote:
> On Wed, 2007-05-30 at 19:34 +0800, jogging song wrote:
> > Hi,
> >        In current x264, deblocking is performed on one row in the
> > encoding loop, rather than on the whole frame after encoding loop.
> > Can this improve the performance? The reason that I can think of is
> > that maybe the data to be used in deblocking
> > is still in cache. Is this right?
> >
> > Best Regards
> > Jogging
>
> Possibly because other threads can use part of the deblocked frame for
> reference before all of it is finished encoding?
> The cache factor probably doesn't hurt, but x264 has a ridiculously low
> percentage of cache misses anyway.
> I believe older versions did frame-at-a-time.

Yes, there is a link to a explanatory diagram in one of the recent posts:

http://forum.doom9.org/showthread.php?t=124557&page=2

-- 
Jindrich Makovicka

-- 
This is the x264-devel mailing-list
To unsubscribe, go to: http://developers.videolan.org/lists.html