Oh, we're not going to be able to do anything that involves copying
stuff around.  I was suggesting that we'd accelerate all or part of
iDCT.

On 5/4/05, Lourens Veen <[EMAIL PROTECTED]> wrote:
> On Wednesday 04 May 2005 02:03, Timothy Miller wrote:
> > Ummm... anyone want to tell me the math that goes into the last MPEG
> > decoding stage just before YUV to RGB conversion?  :)
> 
> That's motion compensation it seems, and the step before that is the iDCT. The
> article at http://www.cs.princeton.edu/research/techreps/TR-516-96 has an
> overview of MPEG-2 decoding in its introduction. There are probably better
> references, but it seems that the last step before conversion essentially
> takes a bunch of frames, and a set of motion vectors and error data, and then
> pieces a frame together by copying blocks from the source frames, shifting
> them according to the motion vector, adding the error data, and pasting them
> into the target frame.
> 
> In practice, I suppose we'd have to support the XvMC extension to be able to
> claim hardware accelerated MPEG playback
> (http://www.xfree86.org/~mvojkovi/XvMC_API.txt).
> 
> Lourens
> 
> 
>

_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)

Reply via email to