[FFmpeg-cvslog] r14687 - in trunk/libavcodec: h264.c h264pred.c h264pred.h

[Michael Niedermayer]

On Mon, Aug 11, 2008 at 07:41:20AM +0200, Diego Biurrun wrote:
> On Mon, Aug 11, 2008 at 04:21:33AM +0200, michael wrote:
> > 
> > Log:
> > Support MBAFF + constrained intra prediction.
> > (no i would not have tried to implement this had i known what mess it is)
> > fixes at least:
> > CAMACI3_Sony_C
> > 
> > --- trunk/libavcodec/h264pred.c	(original)
> > +++ trunk/libavcodec/h264pred.c	Mon Aug 11 04:21:33 2008
> > @@ -690,6 +690,29 @@ static void pred8x8_dc_c(uint8_t *src, i
> >  
> > +//the following 4 function should not be optimized!
> 
> Why?

Because we have just a single file that uses them and that is one of the
conformance streams, not a file found in the wild.
Also even within files that use these functions they are only used when the
left MB pair contains a intra coded MB and an inter coded MB, the current
contains at least one intra coded MB and the left is non interlaced and the
current is interlaced and the current uses DC prediction. Thats quite a
few "and", so it is questionable that these conditions would be true for more
than a few MBs in a file that used MBAFF & constrained intra pred. But until
we see such files in the wild it really makes no sense to optimize ...

[...]

-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Asymptotically faster algorithms should always be preferred if you have
asymptotical amounts of data
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-cvslog/attachments/20080811/4efcb0e8/attachment.pgp>