[XviD-devel] Request: optimized version of image_setedges

[Christoph Lampert]

> 
> How's the status of your SMP aware FrameCodeP function?
> 

The status is exactly "I though of doing it." 

But it's more work than you might think, because there has to be
steps (like there is sync after ME).  So the source would look rather
different than current FrameCodeP and I would have to create copies of 
other routines, too, which I would not really like to do.

I don't think we have to multithread FrameCodeI, because it takes only 
a very tiny part of running time (<2%) and I' not sure about FrameCodeB,
because B-frames can be calculated in parallel anyway (if there are at
least 2 B-frames in a row). 

However, if P is done, B might be easily obtained from that...

At the moment, I won't do B or P (or I), because I don't see much
advantage on dual-CPU PCs at the moment. I would rather be useful for
non-ASM plattforms which is not our main target in my eyes. 

I think, I'll do B-frames ME and better SEARCH8 next. This is really
needed! 



Christoph 


-- 
Christoph H. Lampert chl@math.uni-bonn.de | Diese Signature wurde maschi-     
Beringstr. 6, Raum 14 Tel. (0228) 73-2948 | nell erstellt und bedarf
Sprechstunden: keine, aber meistens da    | keiner Unterschrift. AZ 27B-6