A more interesting thing is that it is WIP on shaders for r300g! Who is working on this?
Announcement
Collapse
No announcement yet.
UVD/hw acceleration If, when?
Collapse
X
-
Originally posted by neuro View Posti just saw the radeon feature matrix:
it appearst that:
"video decode (xvmc/vdpau/va-api) using uvd" became todo. Does that mean that someone gave a green light to uvd or is this just wishfull thinking?
Comment
-
Originally posted by Neuro View PostIt appearst that: "Video Decode (XvMC/VDPAU/VA-API) using UVD" became TODO. Does that mean that someone gave a green light to UVD or is this just wishfull thinking?
Comment
-
Originally posted by Neuro View PostI just saw the Radeon feature matrix:
It appearst that:
"Video Decode (XvMC/VDPAU/VA-API) using UVD" became TODO. Does that mean that someone gave a green light to UVD or is this just wishfull thinking?"TODO" means that someone needs to write the code. The required knowledge to write the code may or may not be known. Please ask on #radeon if you want to get your feet wet on this.
"DONE" = "YES"
"MOSTLY" = "KINDA"
"WIP" = "SOON"
"BIOS" = "DEPENDS"
"N/A" = "CAN'T"
"N/N" = "NOT GONNA/NO POINT"
"TODO" = "NOT YET"
"UNKNOWN" = "DUNNO"
Comment
-
Originally posted by Neuro View PostI just saw the Radeon feature matrix:
It appearst that:
"Video Decode (XvMC/VDPAU/VA-API) using UVD" became TODO. Does that mean that someone gave a green light to UVD or is this just wishfull thinking?
Comment
-
Originally posted by gbeauche View PostBad news for you: "We have not released specs for the video decoder block (UVD) and have no plans to do so". That's from an interview with Terry Makedon, Manager of Software Product Management. Source: http://www.pcinpact.com/dossiers/amd...edon/167-5.htm
In fact logically, "No plans" means nothing more than "has not cleared IP/legal review". It *MUST* clear IP/legal review BEFORE plans can be developed for release.
Further "HAVE PLANS" would logically imply that it HAS cleared IP/legal review, which it has NOT.
Comment
-
Exactly. UVD support is not in the plan right now, but we have said that we will investigate whether it is possible to add it to the plan once we get caught up with existing work (ie we don't plan to put UVD ahead of anything already in the plan).Test signature
Comment
-
Hey, thanks for that link to the article/interview. It was particularly interesting to me. I think it's the same story I suggested before, namely, not enough developers and engineers. ATI/AMD people (here) might insist otherwise or assure everything's okay but that is my perspective and I'm not changing my mind about that any time soon. It's just whether you want to risk enough 'being done' on the Linux side.
I think the quote of 'no plans to do so' refers to no plans to release specs so the UVD features will remain only for the proprietary driver feature side. That's how I interpreted. I notice that no proprietary support is planned for the BSD systems. Like I said, it all seems like not enough manpower and the interviewer was good and persistent by asking about this.
I think I've read a perspective here or somewhere of the hardware being superior to Nvidia's overall but the support seems fragmented (lack of me being able to come up with a better word at the moment!) and undermanned. Too bad too. I don't know if I would want to invest in new or $100+ cards knowing there's so much uncertainty with the drivers. Windows drivers will be priority #1 for a long time and when there's that split proprietary/open source driver project, dare I say you get no optimization on either side because of the divide? Maybe one or the other is good enough, though, but I guess you can't find out unless you try it out with a card? How much is that worth though, that experiment? Ummmm....
Comment
-
Originally posted by bridgman View PostExactly. UVD support is not in the plan right now, but we have said that we will investigate whether it is possible to add it to the plan once we get caught up with existing work (ie we don't plan to put UVD ahead of anything already in the plan).
-- drm: considered done, I guess
-- kms: pretty much done, since it's out of staging
-- r300 and r600 haven't seen too much lately (since the irq and displayport support - almost forgot blitting, but still no tiling)
-- pm: halfway through, but as I see it only Rafal is working on it
So focus might be on:
-- evergreen: I think it's Alex's job (btw, why are you not consistent here and call it r800 in filenames?)
-- pm is still a topic
-- r300g is progressing mostly thanks to Marek and Corbin
What we lack:
-- speed: someone mentioned that the bottleneck is dri2. Is there anyone working on that?
-- r600g: almost nothing so far
-- af: still no specs (is aa implemented?)
-- uvd: still no specs
-- opencl: OK, first we need a proper state tracker in G3D, I guess
-- opengl 3.x: I have no idea if anybody is writing code for this. Maybe GLU3 is a step in this direction.
Regarding all these uvd rants: there was a GSoC project to do this on shaders, right? What about that? Wouldn't that be a relatively easy solution for the time being? Most of us would probably be satisfied with such a partial solution. So, are there any plans to work on that?
You also mentioned the IP reviews. So what are the topics that are underway or queued? Evergreen, af for r600 (at least), uvd. Are there anything else?
Well, I hope I don't seem to be aggressive here. I just collected all the info I could in order to get a clear picture about what is going to happen in the coming months or year. It would be really helpful and much appreciated if you could give me an exhaustive answer. Thank you for your time and efforts!
Comment
Comment