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

[Kostya]

On Sat, Jan 19, 2008 at 03:00:36PM +0100, Christophe GISQUET wrote:
> Hi,
> 
> Kostya a ?crit :
> > I don't like at all.
> > a) exposing function
> > c) checking whether scantable needs transposing in this way is ugly
> > Look how it is done for 8x8 IDCT, for example.
> 
> Well, that's exactly what H.264 does, and Micha?l didn't seem to care.
> 
> You do, but then Micha?l doesn't like my initial proposal, which was
> mimicking what is done for 8x8 idct in intent (except permutation_type
> shouldn't be abused).
> 
> So we have to find a 3rd one. Any idea?

make a function to permutate permutation tables and merge it with
vc1dsp_init*()

But first of all, can you provide a benchmark for real speedup
when using MMX version of 4x4 transform? Preferably on several samples
from samples.mplayerhq.hu, big and small.
 
> > b) doxygen comment should belong to another patch
> 
> Fair enough.
> 
> Best regards,
> -- 
> Christophe GISQUET