[MPlayer-dev-eng] optimized colorspace for ALTIVEC enabled platforms

Romain Dolbeau dolbeau at irisa.fr
Thu Mar 11 08:29:27 CET 2004


Marc Hoffman wrote:

> So I guess i won't get the opportunity to work on video scaling next, 
> hehe.  Well excellent, you got that covered.

Not fully. I got bitten by Amdahl's Law before :-) I have only
seen yuv2yuvX be a cpu-hog, so I only did this one.

 > We probably might want to
> think about how the two modules can work together, it appears that the 
> capabilites are tightly coupled in the postproc module.  If you have any 
> ideas shoot them at me.  I've been playing with someother formats 
> currently not supported by my patch and just added them in and realized 
> that the current structure is a maintenance nightmare.  So I'm thinking 
> about how to best structure the core algorithm so that we can reuse it 
> with multiple output and input formats a bit better.

Trying to "fix" or "improve" the structure isn't going to be
very productive IMHO. It ain't nice but it get the job done,
and once you understand it it's not *that* bad. And the
more sophisticated the function you change, the more likely
it will need to be redone because of change in the C version.
That's why I usually keep myself to only DSP-like functions.

-- 
Romain Dolbeau




More information about the MPlayer-dev-eng mailing list