Results 1 to 9 of 9

Thread: Nouveau Hardware-Based Video Support In Mesa

  1. #1
    Join Date
    Jan 2007
    Posts
    14,810

    Default Nouveau Hardware-Based Video Support In Mesa

    Phoronix: Nouveau Hardware-Based Video Support In Mesa

    Due to being busy with XDC2011 Chicago, there was some news missed over the weekend: there's the Mesa user-space side support for Nouveau video decoding using a hardware-based NVIDIA video decoder...

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

  2. #2
    Join Date
    Jan 2008
    Posts
    138

    Default Who watches MPEG-1 videos nowdays?

    Where do you find MPEG-1 clips nowdays?

  3. #3
    Join Date
    Apr 2011
    Posts
    12

    Default

    Probably nowhere, but MPEG-2 (also supported) is still used on DVDs and digital TV. If you can find any card between a Geforce FX and a Fermi (some excluded) you might just have a little more CPU left for other tasks (such as recording digital TV signals?), a welcome addition for computers that are so old that they actually have such a card I'd say.
    Last edited by RSpliet; 09-13-2011 at 05:10 PM.

  4. #4
    Join Date
    Dec 2008
    Posts
    988

    Default

    Quote Originally Posted by RSpliet View Post
    Probably nowhere, but MPEG-2 (also supported) is still used on DVDs and digital TV. If you can find any card between a Geforce FX and a Fermi (some excluded) you might just have a little more CPU left for other tasks (such as recording digital TV signals?), a welcome addition for computers that are so old that they actually have such a card I'd say.
    Recording digital TV is little more than reading the transport stream from the digital TV receiver and writing it to the hard disk, and maybe some PID filtering if this isn't implemented in the hardware. It hardly takes any CPU cycles.

    Having MPEG-2 decode acceleration may be interesting for maybe 1 percent of Linux users, or even less, but I think it would be accurate to say that the majority of people who are waiting for this Gallium based video decode acceleration are waiting for H.264 support.

    Still, those nouveau hackers are really amazing in what they are doing. I wouldn't be surprised if they have H.264 support in their driver before ATI has.

  5. #5
    Join Date
    Sep 2007
    Location
    Connecticut,USA
    Posts
    962

    Default

    Quote Originally Posted by monraaf View Post
    Recording digital TV is little more than reading the transport stream from the digital TV receiver and writing it to the hard disk, and maybe some PID filtering if this isn't implemented in the hardware. It hardly takes any CPU cycles.

    Having MPEG-2 decode acceleration may be interesting for maybe 1 percent of Linux users, or even less, but I think it would be accurate to say that the majority of people who are waiting for this Gallium based video decode acceleration are waiting for H.264 support.

    Still, those nouveau hackers are really amazing in what they are doing. I wouldn't be surprised if they have H.264 support in their driver before ATI has.
    Wouldn't then ffmpeg be able to support this HW decoder? That would be awesome once this HW decoder code matures! Definitely useful progress has been made on the nouveau front lately and there's hope that the driver would be on a par with the nvidia blob as well as being ahead of the radeon driver sometime considering that this new work is a big step forward in that regard.

  6. #6

    Default

    Quote Originally Posted by phoronix View Post
    Phoronix: Nouveau Hardware-Based Video Support In Mesa

    Due to being busy with XDC2011 Chicago, there was some news missed over the weekend: there's the Mesa user-space side support for Nouveau video decoding using a hardware-based NVIDIA video decoder...

    http://www.phoronix.com/vr.php?view=OTkwMw
    Are you sure that it is not the PureVideo decoder?

    http://en.wikipedia.org/wiki/Nvidia_PureVideo

    Nvidia is using that, but they only use the earlier generations on Windows.

  7. #7
    Join Date
    Jan 2010
    Posts
    364

    Default

    I don't really see the point of this. MPEG-2 is very cheap to decode, even the slowest CPUs you can find nowadays are able to decode high-bitrate HD resolution MPEG-2 video. If this work isn't going to lead to H.264 acceleration in the end, it's useless.

  8. #8
    Join Date
    Jun 2006
    Location
    Portugal
    Posts
    527

    Default

    Quote Originally Posted by brent View Post
    I don't really see the point of this. MPEG-2 is very cheap to decode, even the slowest CPUs you can find nowadays are able to decode high-bitrate HD resolution MPEG-2 video. If this work isn't going to lead to H.264 acceleration in the end, it's useless.
    I think every improvement is welcome. For example, if you look at the mailing list, this work also prompted some discussion on the mesa api's for video decoding, and what might be changed.

    Even if this doesn't directly impact h264 decoding, it might get the APIs and other stuff in better order for the future. Or not. But why complain about a feature that increases performance with no downsides?

  9. #9
    Join Date
    Oct 2009
    Posts
    2,110

    Default

    This is actually quite fascinating.

    To all... yes, these blocks would definitely be the same video decoders as the purevideo/vdpau blocks.
    These blocks should be present in any geforce 5 series or newer, although growing in capability with each incarnation.

    Geforce 5's... NV3x cores, had MPEG1/2 decode.
    Geforce 6's... NV4x cores, had improved deint and overlay resizing compared to the NV3x.
    Basic h264 acceleration (limited to MC) was added to the NV43 core.
    Press releases suggested that it may be present (but disabled) as far back as NV40 (early geforce 68). It *may* be possible to access the h264 MC block in all geforce 6's. The question of course, is IF they are present, why were they disabled? Answer may be that they are defective.

    All geforce 6's except some early 68's have at least an NV43. That includes all ENTRY LEVEL geforce 6's -- the 6100 and 6150 (many laptops) are NV44's and so have h264 MC blocks. Interestingly -- note that I don't know how accurate this is -- the 6150 is said to have "HD video acceleration" (wikipedia), while the 6100 does not. What is interesting about this is that the 6100 and 6150 are the exact same chip, clocked differently, and evidently have certain features enabled or disabled based on the model number (most likely by software). I would suspect that the h264 MC could be enabled on the 6100.

    Geforce 7's, being more or less the same as geforce 6's, appear to have the same video decode blocks as NV43+... with the exception of the "M"'s (laptops), which have "VDPAU feature set B"... supposedly.

    Geforce 8's and later, have "VDPAU feature set A" or better.


    So... this could be *really* good news for anyone with a geforce 6 or 7, with the *possible* exception of anything below NV43, and the 7-M's (which possibly have full blown VDPAU). I, for one, wouldn't mind it if my crusty old C51 laptop could manage to decode an h264.

Posting Permissions

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