On Tue, 29 Aug 2006 13:48:13 -0600
"Lance Hanlen" <[EMAIL PROTECTED]> wrote:
> I'm just getting up to speed on it now, but other than the YUV<->RGB
> conversion Tim and I are looking at, my experience so far is that
> pretty much every (in use) video codec has the basic JPEG alphabet:
> Huffman, entropy, quantization, iDCT, et. al. I found at least some of
> this stuff on opencores...
Yes, of course. Most of the codecs out there are 8x8 block
DCT based, which asks for a motion compensation to be efficient,
some entropy coding etc. Clearly, you can put a lot of these
functions into a card to accelerate decoding, but the problem
is that either you have to be able to do full decoding from
some point in the process on until you have a picture on the
card, or you have to transfere back the data into host memory
which makes the whole acceleration useless.
The problem here lies in the way how the building blocks
of the codec are combined and how the data is stored.
You would need to be able to combine all the blocks needed
and have the data in the right format to gain anything.
Yes, it can be done. Yes, it was done before. But i do
not think that the it is worth the effort for our project.
Attila Kinali
--
心をこめて聞け心をこめて話せ
_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)
