[FFmpeg-devel] [RFC] [PATCH] Indicate better when transposing zigzag scantables is needed in some codecs

Kostya kostya.shishkov
Sat Jan 19 14:37:19 CET 2008


On Sat, Jan 19, 2008 at 01:13:30PM +0100, Christophe GISQUET wrote:
> Christophe GISQUET a ?crit :
> > pseudo-code would be:
> > if (idct!=idct_c && idct!=idct_ppc) // or (idct==idct_mmx)
> >   transpose4x4;
> 
> In that case, as there is no 4x4 idct for ppc, this approach can be used.
> 
> I haven't noticed any BOFH in a 50m radius,

You are not supposed to notice them, until it's too late.

> so I expect the attached
> patch to change the code so that it performs a conditional transpose of
> the 4x4 zz scantable. I've decided to split it from the mmx patch
> because it belonged more to that topic, although it does nothing by
> itself and is a prerequisite for the mmx patch.
> 
> I can also add it back as a 2nd patch along the mmx patch over the other
> thread if that's preferred.

I don't like at all.
a) exposing function
b) doxygen comment should belong to another patch
c) checking whether scantable needs transposing in this way is ugly
Look how it is done for 8x8 IDCT, for example.

And I suspect all of this would be like shaving piglets - much squeal with
no significant benefit.




More information about the ffmpeg-devel mailing list