Page 1 of 2 12 LastLast
Results 1 to 10 of 14

Thread: Using GPU Video Acceleration Under Linux

  1. #1
    Join Date
    Jan 2007
    Posts
    15,650

    Default Using GPU Video Acceleration Under Linux

    Phoronix: Using GPU Video Acceleration Under Linux

    For those intrigued by the article written by XBMC developers after being frustrated by AMD's Catalyst Linux driver, here's a look at some of the other options for GPU-based video acceleration under Linux...

    http://www.phoronix.com/vr.php?view=MTEyNDk

  2. #2
    Join Date
    Jan 2009
    Posts
    1,766

    Default

    off topic a bit and maybe stupid question but anyway:

    will someone with a ivy bridge cpu/gpu and another card using open drivers be able to use ivy for VA and the other card for the rest once the multi gpu work by dave arlie lands??

  3. #3
    Join Date
    Feb 2012
    Posts
    515

    Default

    Would've been nice if you also went further than just decoding. For example, VAAPI currently has only basic bob deinterlacing. And there's no deinterlacing of mpeg2 video on Ironlake, because of colorspace issues: http://lists.freedesktop.org/archive...ry/014847.html

    On the other hand, VDPAU has motion-adaptive spacial-temporal deinterlacing (yay buzzwords ), both single-rate and double-rate. The double-rate one is excellent for watching live sports, like the currently ongoing Euro2012 . Then there's IVTC (inverse telecine).

    No idea about deinterlacing on XvBA.


    Further, both VDPAU and VAAPI provide high-quality scaling. VDPAU also has sharpening and denoising, though I haven't tested those two yet.
    Last edited by Gusar; 06-22-2012 at 04:31 PM.

  4. #4
    Join Date
    May 2007
    Location
    Third Rock from the Sun
    Posts
    6,587

    Default

    VDPAU support is now widespread throughout the popular programs like MythTV, XBMC, VLC, MPlayer, etc.
    Correct me if I'm wrong, but I believe that VLC only supports vdpau via VA-API.


    The S3 Graphics driver even claims VDPAU support although I don't have any modern discrete S3 graphics card nor do I actually know anyone with such hardware to confirm this claim.
    Worked fine on a S3 GT 530 a year or so ago that I had access too.

  5. #5
    Join Date
    Feb 2011
    Posts
    163

    Default

    Xvba das have internal deinterlacing. It is BOB Quality - so nothing compared to what you get out of vdpau. For a comparison, here are the missing bits: http://ati.cchtml.com/show_bug.cgi?id=448

  6. #6
    Join Date
    Sep 2008
    Location
    Vilnius, Lithuania
    Posts
    2,666

    Default

    Quote Originally Posted by 89c51 View Post
    off topic a bit and maybe stupid question but anyway:

    will someone with a ivy bridge cpu/gpu and another card using open drivers be able to use ivy for VA and the other card for the rest once the multi gpu work by dave arlie lands??
    Yea, I had a very similar idea after reading this... If so, it would be wonderful and getting an integrated video chip might be worth it after all.
    Last edited by GreatEmerald; 06-22-2012 at 06:23 PM.

  7. #7
    Join Date
    Sep 2008
    Location
    Vilnius, Lithuania
    Posts
    2,666

    Default

    I forgot to mention, good work on this article, Michael, it was very informative! I never really realised what each of the vendors' solutions used for the video acceleration, and this makes things quite clear!

  8. #8
    Join Date
    Oct 2009
    Posts
    101

    Default

    I run XBMC on an AMD E-350 APU. The CPU is not nearly powerful enough to decode HD video in real-time, so XvBA is the only real option. First, I tried mainline XBMC with the XvBA/VA-API bridge. Crash city. Fortunately, I found a branch of XBMC that uses XvBA directly. This one doesn't crash, but pausing and then resuming playback results in a few seconds of A/V desync. AMD really needs to stop screwing around and just implement VA-API or VDPAU directly (and not botch it).

    The thing that really pisses me off, though, is that supporting proper video decoding in the open-source drivers will require reverse-engineering because AMD can't release the specifications because of WINDOWS DRM. Fuck you, Microsoft, and fuck you, AMD, for capitulating.

  9. #9
    Join Date
    Feb 2008
    Posts
    214

    Default

    This is a mess...

    So much APIs for the same task. Why not merging them? Fuck all this mess done by cops, they want to annoy the others and they annoy the users and developers even more. Why not just use OpenMAX instead?

    Seriously, most of the Linux problems are due that most important projects lack of proper leadership and strong management. Xorg/MESA should have a stronger presence in Khronos Group, for example


    Seriously, Xorg/MESA should be a lot stronger organization with a stricter leadership.

    Michael: What about an article doing a nice comparison between all those video acceleration APIs? You have some interesting bits in the comments of this article

  10. #10
    Join Date
    Sep 2011
    Posts
    709

    Default

    One some Nvidia cards you end up with video having a blue tint, I don't rember if this was fixable with a setting or you had to switch VA off.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •