Adobe Flash 10.2 Brings Linux Video Acceleration

[rocco]

yay, got 32-bit libvdpau and acceleration is working, but still, 2 questions:
1. How this multi-monitor mode should be working? I got primary monitor 1680x1050 and secondary 1920x1080 and going full screen on youtube on secondary monitor makes it first monitor size and the rest is black.
2. Clicking on secondary monitor makes full screen turn off.
I'm using twinview.

[deanjo]

Not sure, you might have to have your displays running at the same resolution first.

[rocco]

Changing resolution is not a solution, I checked with bigger screen set as primary and fullscreen is really fullscreen, but on secondary monitor video is cropped, so I don't see:

proper full-screen support in multi-monitor configurations, and native support for custom mouse cursors

Or maybe I'm missing something? Like explicit resolution in xorg.conf instead of nvidia-auto-select?

[deanjo]

Anyways here is what the CPU usage looks like on a 8200 IGP board with a AMD 4200+ locked in at 600 Mhz.

[deanjo]

Changing resolution is not a solution, I checked with bigger screen set as primary and fullscreen is really fullscreen, but on secondary monitor video is cropped, so I don't see:

Or maybe I'm missing something? Like explicit resolution in xorg.conf instead of nvidia-auto-select?

Just throwing out ideas, what Adobe considers "proper full-screen support in multi-monitor configurations" I have no idea. This is adobe we are talking about after all and they have a different definition of what "proper" is a lot of the time.

[rocco]

OK, did some more testing. Multimonitor mode works just fine with firefox. With chromium full screen seems to be the size of monitor on which chromium was started. Can somebody confirm? And one previously mentioned - is there any way of not turing off full screen when clicking on secondary screen?

[Michael]

Netbook results for this Flash video acceleration on a netbook being published tomorrow to show the differences when it comes to CPU usage / GPU usage / power consumption / etc.

[gbeauche]

or direct xvba support, but then ati really has to fix the drivers first for hd 5, as the hack is only in the xvba-video wrapper...

I re-opened the hd5 bug many weeks ago mentioning explicitly this Flash work. Yes, it was closed as "won't fix". However, the bug is now scheduled to be *reproduced* around 2011. i.e. no schedule to get it fixed yet. Though, someone kind enough in engineering may probably look at it in his spare time and get the thing fixed soon? Who knows... Besides, two other bug fixes (well, one bug fix and one very small feature impl, which is almost already implemented) are needed, AFAICS.

BTW, may an hd 6 user be kind enough to check the xvba-video workaround works for him?

[brent]

They're both designed for decode acceleration that's integrated with the GPU, and CrystalHD isn't. I don't think it could even support either of the APIs without video driver-specific magic.

The decoding parts of VA-API and VDPAU map just fine to CrystalHD and would provide the same functionality without introducing yet another API. The presentation parts obviously don't apply to CrystalHD. However, a generic OpenGL-based implementation is possible. No "magic" is needed in any case.

[djtm]

Well vdpau or not, this flash player version is much faster on my notebook (which has neither vdpau nor vaapi support, only xvmc)