AMD Radeon VDPAU Video Performance With Gallium3D

[dungeon]

The answer to which of those questions you quoted? Come on, only one time give us a clear and precise answer. If there is progress, what for progress?

I bet on first .

And sadly every time you show up to comment on that topic you fail to actually answer the question: Has there been any progress made on UVD for RS780/880 in the last, lets say 9 months? If not, would you still call this low priority ? Or would you call it what it is: Nobody is working on that anymore?

Second and third question depends on No answer, but because answer was Yes there can't be logical answer on those .

[bridgman]

I bet on first

Ladies and gentlemen, we have a winner !!

The answer to which of those questions you quoted? Come on, only one time give us a clear and precise answer. If there is progress, what for progress?

Your text I quoted talked about "the question:". That one.

Since the whole review/sanitize/release exercise is about determining what secrets are safe to expose, I can't really go into more detail. You know this from the last dozen times we've been through a complex IP review.

[Vim_User]

Since the whole review/sanitize/release exercise is about determining what secrets are safe to expose, I can't really go into more detail. You know this from the last dozen times we've been through a complex IP review.

I have not asked you to reveal company or trade secrets, I have asked you for a precise answer about the progress of this procedure. Can you at least give an answer about how far this procedure has already gone? For sure legal team/developers will not discuss the same thing over and over again. There must be some kind of "this has already clearance for release" and "this has to be reviewed further" list. If you could at least give us some numbers on items on each list that would already help to give at least some confidence that actually something is happening(especially if we ask in half a year again, so that we can see if there really is progress).

[profoundWHALE]

Vim_User, how do you reconcile all the work that's been done (and is still being done) to support your hardware in the open source drivers with the statements you're making ? You're the one who said we're not going to do it, not us -- and AFAICS you're acting as if *we* said we wouldn't ever release it.

What we said is that yes, we have done a lot of work on it, yes there's been a lot of progress, yes there is still some work to do before we can release, yes we're still working on it, and like EVERY OTHER SCRAP OF OPEN SOURCE SUPPORT WE HAVE RELEASED IN THE LAST SEVEN YEARS we won't know until we're finished if & when we will be able to release it.

The answer to which of those questions you quoted? Come on, only one time give us a clear and precise answer. If there is progress, what for progress?

http://www.phoronix.com/forums/showt...465#post420465

And now Vim_User is still going at it. I don't see where the confusion is but this thread is starting to sound like a broken record.

[bridgman]

I have not asked you to reveal company or trade secrets, I have asked you for a precise answer about the progress of this procedure. Can you at least give an answer about how far this procedure has already gone? For sure legal team/developers will not discuss the same thing over and over again. There must be some kind of "this has already clearance for release" and "this has to be reviewed further" list. If you could at least give us some numbers on items on each list that would already help to give at least some confidence that actually something is happening(especially if we ask in half a year again, so that we can see if there really is progress).

OK, I'm starting to think you have an out-of-whack view of how the review/revise/release process happens. It's not like there's a single deterministic list of tasks defined at the start that get ticked off one at a time, more of an iterative process that keeps looping around until we find a solution that can be released. We've already been through the loop more than once but there's no magic number of attempts where someone feels sorry for us and we get a free pass.

In some cases the solution is to not expose something we already have approval for in order to be able to expose something else, so telling you the parts that have been approved runs the risk of not ever being able to release in the case where we need to hold one of them back.

[dungeon]

Ladies and gentlemen, we have a winner !!

Thanks, dungeon giving the prize to Vim_User .

[Kivada]

Youtube uses h246 for most of its videos. I think it falls back to webm if your browser doesn't support h246. And gstreamer with HWA crashes firefox in all systems.


Also there is one more thing with vdpau. If you play a video with gst-play and try to run a vlc video with vdpau enabled it will not play it. +vlc uses more cpu than gst-play.

They offer almost every video in .webm in the "medium" resolution of 640x360 even if the video say's it's flash or .mp4 only. just use the FlashGot extension on any Youtube vid to see all available versions of the video. Said videos will also almost always play just fine if embedded in an outside webpage but with complain about not being compatible on the actual Youtube site.

[Ansla]

OK, I'm starting to think you have an out-of-whack view of how the review/revise/release process happens. It's not like there's a single deterministic list of tasks defined at the start that get ticked off one at a time, more of an iterative process that keeps looping around until we find a solution that can be released. We've already been through the loop more than once but there's no magic number of attempts where someone feels sorry for us and we get a free pass.

In some cases the solution is to not expose something we already have approval for in order to be able to expose something else, so telling you the parts that have been approved runs the risk of not ever being able to release in the case where we need to hold one of them back.

I think he expects you to tell him something like "we went through 2 more iterations since last time you asked"

[Ansla]

Power usage

Hope nobody minds if I go a bit on topic on this old thread, but I think the article missed the essential about decode offloading. Lowering the CPU usage is a nice side-effect, but not the goal itself. The real deal with decode offload is being more power efficient. So how much power you save by using VDPAU? Here is a benchmark that will answer just that: http://openbenchmarking.org/result/1...SO-U2708169054

A little sumary:
- idle ~12W
- GL/XV ~18W (6W more than idle)
- VDPAU ~15W (3W more than idle)

So it looks like UVD uses about half the power compared to the CPU for decoding the Phoronix sample on the E-450 APU.

[dungeon]

Hope nobody minds if I go a bit on topic on this old thread...

It is not that old, just one week .

little sumary:
- idle ~12W
- GL/XV ~18W (6W more than idle)
- VDPAU ~15W (3W more than idle)

So it looks like UVD uses about half the power compared to the CPU for decoding the Phoronix sample on the E-450 APU.

Yes that was for this sample, but if you use another sample:

http://www.phoronix.com/forums/showthread.php?102776-AMD-Radeon-VDPAU-Video-Performance-With-Gallium3D&p=426643#post426643

Which is more CPU resource hungry, VDPAU will stay on his 3W, so for that one playing with Xv/GL i guess will consume 12W... end result would be 4 times more power consumpation .

edit: actually you can't play that 60 fps sample with Xv/GL on E-450, could you? But with vdpau you can ?