AMD's UVD2 & XvMC For Linux?

[jeffro-tull]

@ cruiseoveride

I was also curious about that sentence. And I share the same sentiment (though not to the same extent) about all the hush-hush in articles I read on this site. It does get kind of old seeing "well, we can't talk about that just yet" or "now we'll speak up" and things of that nature.

But, as a linux user with an R500, it's either dig through mailing lists, press releases, and IRC logs or put up with Phoronix's occasional secrecy. Laziness wins out.

[dashcloud]

Ok I will then.



So lets break this sentence down and see what we can find.

"With the information...", this is referring to the previous statement:


"..out there", obviously referring to the public domain/internet.

", now we're willing to talk a bit more" hmm

This is strange, because if: phoronix/Michael was under an NDA over ATI UVD/Linux/Whatever, he would not be able to "talk" about it, until the embargo was lifted. But since he says "now we're willing" it gets tricky, why "now", did you have to wait for someone else to report so-called classified information, before you could? but how can you in the first place if you were under an NDA, already exposed information or not?

I'm pretty sure Michael's NDA has a clause which says if information under the NDA gets out due to no fault of your own, you can then speak to that information only (or they got the OK from AMD)

[Uber]

This is great news!
Will my X1250 with Avivo but no UVD, also benefit from this?

AIK UVD won't be supported anyhow due to the antifunctions.

If the spechs for the antifunctions where known you could actualy write drivers that makes things that "just works".

AIK UVD2 will be supported in the future.

But XVMC would be an very welcome gesture for UVD, even with the usual ATI tearing ;D

There is an thread where video acc and antifunctions are discussed.

[W3ird_N3rd]

What will be interesting though is how AMD decides to implement their high-definition video support on Linux. Seeing as there is no suitable standard right now, they will likely introduce their own interface.

How about VAAPI? Intel is going to support it (already does on some ATOM chipsets I understood), if AMD would support it as well I guess ffmpeg support wouldn't be far away and that would bring support to mplayer, xine etc. And at that point nVidia won't really have any choice.

ATM there is no standard to accelerate h.264 (XvMC only supports MPEG2), VAAPI should be it but seems not to be used..

[mtippett]

How about VAAPI? Intel is going to support it (already does on some ATOM chipsets I understood), if AMD would support it as well I guess ffmpeg support wouldn't be far away and that would bring support to mplayer, xine etc. And at that point nVidia won't really have any choice.

ATM there is no standard to accelerate h.264 (XvMC only supports MPEG2), VAAPI should be it but seems not to be used..

Which interface are you talking about? Intel is developing two competing ones - VAAPI and an extension to XvMC.

[W3ird_N3rd]

Which interface are you talking about? Intel is developing two competing ones - VAAPI and an extension to XvMC.

I can't post links at this moment I believe (or my post would have to be reviewed), but in the freedesktop wiki there is an article Software/vaapi. I have already seen libva (an implementation of VAAPI) packages for Ubuntu, so all that should (apart from support in ffmpeg) pretty much already exist, or am I mistaken?

That freedesktop article says:

The main motivation for this proposal is to enable hardware accelerated video decode at various entry-points (VLD, IDCT, Motion Compensation etc.) for the prevailing coding standards today (MPEG-2, MPEG-4 ASP/H.263, MPEG-4 AVC/H.264, and VC-1/VMW3). Extending XvMC was considered, but due to its original design for MPEG-2 MotionComp only, it made more sense to design an interface from scratch that can fully expose the video decode capabilities in today's GPUs.

So I do not believe Intel works on extending XvMC.

[mtippett]

I can't post links at this moment I believe (or my post would have to be reviewed), but in the freedesktop wiki there is an article Software/vaapi. I have already seen libva (an implementation of VAAPI) packages for Ubuntu, so all that should (apart from support in ffmpeg) pretty much already exist, or am I mistaken?

That freedesktop article says:

So I do not believe Intel works on extending XvMC.

Intel - Keith Packard's team talking about XvMC extensions - http://www.x.org/wiki/Events/XDS2007...cb82cad622a148

Intel - Jonathon Bian is the prime contact for VAAPI - http://freedesktop.org/wiki/Software/vaapi


I don't know the interplay between them, but it looks like Jonathon was one of the original authors of the i810 driver and Keith is the new kid on the block (Intel driver wise).

But it is clear that *BOTH* efforts are being funded by Intel, and *BOTH* seem to want to ignore the other. A Wiki article is hardly ever authoratitive or accurate .

Regards,

Matthew

[meden]

AIK UVD won't be supported anyhow due to the antifunctions.

If the spechs for the antifunctions where known you could actualy write drivers that makes things that "just works".

AIK UVD2 will be supported in the future.

But XVMC would be an very welcome gesture for UVD, even with the usual ATI tearing ;D

There is an thread where video acc and antifunctions are discussed.

Can someone "at high floors" (say... Bridgman) confirm this, so that I can stop to be happy (along with my HD2600)?

[bridgman]

My position in this has been very consistent -- "I don't know yet"

Once we get 6xx/7xx 3D engine info released to the public and (hopefully) some basic power management stuff, video is next on the list for us to look at. Right now I have high confidence that IDCT (HD3xxx and earlier) will be released, low confidence that UVD (HD2xxx/HD3xxx) will be released, and slightly higher confidence that UVD2 (rs780, HD4xxx) will be released -- but won't be sure on any of them until we go through the risk & IP analysis effort.

Whether support for that hardware is implemented with XvMC or VAAPI or something else is up to the community. If our developers implemented the support first we would probably go with XvMC and try to align extensions with Intel to make things easier for player developers.

Use of video acceleration hardware by the proprietary driver is a different story -- we have a bit more freedom to use the hardware in fglrx without the associated IP hassles, but only to the extent that we can discourage reverse engineering or at least publication (say, in an open source driver) of the resulting information.

[TechMage89]

Brig Bridgman, I'm not really sure from an IP/ DRM perspective that supporting video decode in the Linux Catalyst is any riskier than in the Windows version. I'm guessing that it's probably not much harder for a determined hacker to RE the Windows version