[mythtvnz] State of h264 decoding

[David Zanetti]

On Mon, 2009-01-05 at 09:59 +1300, Bernard Mentink wrote:
> No you did not. I did not mention an nvidia card. I was talking about
> stability of software decoding of h264 ...

You may wish to just tone things down a bit, as you'll find you get
better support from mailing lists if you don't bite people's heads off
when they read this in your post:

"Also any word on hardware decoding support in the kernel?"

Of which the answer is nvidia have running code and working drivers,
today, everyone else (Intel, ATI, whatever) have either nothing or only
suggested APIs. I don't think it's unreasonable when you're asking about
*hardware* decode to mention what hardware is good.

The kernel features being mainlined in 2.6.28 are only involved in
allowing the kernel to set graphics modes (ie, x*y at nbbp) and provide
a common memory manager for GPUs whose drivers choose to use it.

There are no plans to mainline a generic kernel h264 acceleration API,
and I doubt one is likely to be proposed.

G45 support for GPU accel is nonexistent and, if I'm reading the
datasheets right, limited even if they get it going. There appear a
whole slew of things it won't do.

As for future ffmpeg (software) decode, I think people will end up
spending a lot less time and effort on it with a clean dependable set of
APIs to do it in hardware, so I would actually expect software decode
improvements to slow a fair bit.

-- 
David Zanetti <dave2 at wetstring.net>
http://hairy.geek.nz/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
Url : http://lists.ourshack.com/pipermail/mythtvnz/attachments/20090105/bc84be62/attachment.pgp