Page 1 of 4 123 ... LastLast
Results 1 to 10 of 38

Thread: Nouveau In Linux 2.6.36 Has NVIDIA Fermi Mode-Setting

Hybrid View

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

    Default Nouveau In Linux 2.6.36 Has NVIDIA Fermi Mode-Setting

    Phoronix: Nouveau In Linux 2.6.36 Has NVIDIA Fermi Mode-Setting

    Last week we reported on the first DRM pull request for the Linux 2.6.36 kernel and it brought forth some interesting features and improvements for the open-source Linux graphics drivers living within the kernel, but a second pull request has reached the email inbox of Linus Torvalds this afternoon...

    http://www.phoronix.com/vr.php?view=ODQ5MQ

  2. #2
    Join Date
    May 2009
    Location
    Florianopolis, Brazil
    Posts
    54

    Default

    See what one can do if one gives a determined hacker some hardware, sure it would be faster with specs and assistance from the engineers, but this is still pretty impressive.

  3. #3
    Join Date
    Aug 2007
    Posts
    437

    Default

    The fact is, this reverse engineered hacking driver still progresses faster than the opposite party who has supposedly 'FULL ASSISTANCE AND DOCUMENTATION', pretty ironic isn't it? Good job Nouveau devs! You deserve the tip of hat

  4. #4
    Join Date
    May 2007
    Posts
    319

    Default

    Quote Originally Posted by FunkyRider View Post
    The fact is, this reverse engineered hacking driver still progresses faster than the opposite party who has supposedly 'FULL ASSISTANCE AND DOCUMENTATION', pretty ironic isn't it? Good job Nouveau devs! You deserve the tip of hat
    /me wonders which hw is the opposite party and which opposite party hw is missing modesetting.

    Or is FunkyRider just a bit of a flaming idiot.

  5. #5
    Join Date
    Sep 2008
    Posts
    270

    Default

    Quote Originally Posted by airlied View Post
    /me wonders which hw is the opposite party and which opposite party hw is missing modesetting.

    Or is FunkyRider just a bit of a flaming idiot.
    pwnd!

    Oh, and when will we finally be able to edit again and have posts of less then ten characters?

  6. #6
    Join Date
    Oct 2008
    Posts
    3,246

    Default

    Quote Originally Posted by MaestroMaus View Post
    pwnd!

    Oh, and when will we finally be able to edit again and have posts of less then ten characters?
    Never? I seem to remember when it started it was supposed to be a temporary measure, but obviously Michael has decided not to fix it.

  7. #7
    Join Date
    Jul 2007
    Posts
    176

    Default

    Hi Michael, I hope I am not nitpicking, you mentioned
    ... Freescale offered free Fermi GPUs to open-source developers ...
    , but I believe it is PathScale, and I couldn't verify that they are related to Freescale in any way (or may be I am missing something). If so, a correction (and due credit to PathScale) would be helpful.

    Back on topic, I have mentioned this previously before in Phoronix forums, but repeating. As much I am thankful to Nvidia for their fully functional proprietary drivers, I am grateful beyond words to Nouveau developers. I was super surprised when I tried Nouveau recently and I saw accelerated text terminals, flawless mode switching, randr 1.2 screen management, working 3d (and 2d) and kwin compositing effects and a overall functional driver. Currently sticking to proprietary drivers for power management (as I have a heavy duty 9800m gs in my notebook) and gpu computing using cuda and opencl.

  8. #8
    Join Date
    Nov 2008
    Posts
    784

    Default

    Evergreen / radeon:
    release: sept/oct 09
    modesetting (KMS and UMS): beginning of February, ~4 months

    Fermi / nouveau:
    release: Nov 09. Huang said so. Ok, let's make that april 2010.
    modesetting (KMS only): now. ~4 months.


    AMD spent additional time on IP review and they wrote extensive documentation while they were at it. Their task was more difficult this time since they had to deal with huge changes in the display logic due to Eyefinity.
    Nouveau spent additional time RE'ing the cards. Their task was easier since they skipped writing UMS code and the display logic should be similar to the older cards. They can also blindly copy what the blob does, they're not required to understand and document it.


    So as we can clearly see, both apples and oranges took around 4 months to develop.

  9. #9
    Join Date
    Oct 2009
    Posts
    2,145

    Default

    Quote Originally Posted by rohcQaH View Post
    Evergreen / radeon:
    release: sept/oct 09
    modesetting (KMS and UMS): beginning of February, ~4 months

    Fermi / nouveau:
    release: Nov 09. Huang said so. Ok, let's make that april 2010.
    modesetting (KMS only): now. ~4 months.


    AMD spent additional time on IP review and they wrote extensive documentation while they were at it. Their task was more difficult this time since they had to deal with huge changes in the display logic due to Eyefinity.
    Nouveau spent additional time RE'ing the cards. Their task was easier since they skipped writing UMS code and the display logic should be similar to the older cards. They can also blindly copy what the blob does, they're not required to understand and document it.


    So as we can clearly see, both apples and oranges took around 4 months to develop.
    The clear advantage of the radeon development path is when it comes time to create new capabilities that aren't part of the blob driver. Having spec means that you understand what things do, which makes this MUCH easier. If the nouveau developers really need to understand what the thing does, they may be up a creek.

    Note also that the nouveau developers have neither need nor interest in protecting nvidia IP. From this perspective, nvidia is really shooting themselves in the foot with this, because it means that some of the proprietary secret sauce will get out of their control. That could have SERIOUS ramifications for them down the road...

  10. #10
    Join Date
    Sep 2007
    Location
    Connecticut,USA
    Posts
    983

    Default

    Quote Originally Posted by droidhacker View Post
    The clear advantage of the radeon development path is when it comes time to create new capabilities that aren't part of the blob driver. Having spec means that you understand what things do, which makes this MUCH easier. If the nouveau developers really need to understand what the thing does, they may be up a creek.

    Note also that the nouveau developers have neither need nor interest in protecting nvidia IP. From this perspective, nvidia is really shooting themselves in the foot with this, because it means that some of the proprietary secret sauce will get out of their control. That could have SERIOUS ramifications for them down the road...
    That's a good point. Without any docs or any assistance from nVidia, the nouveau devs won't be able to add any other features as they don't have any clear way to add that feature.

    As for the nouveau devs getting close to the "secret sauce", I am sure there may be some noise from nVidia if the feature and performance parity got quite close to the reference nvidia blob. It could be well within nVidia's right to demand a code audit in that case for them to determine if any of their patents or IP got infringed. I'm sure that as long as the devs did their work totally in a clean room with NO access to docs or code of nVidia's that there is no case for patent infringement if the nouveau driver reaches parity with nvidia's reference blob in terms of features and performance (I am not a lawyer)

Posting Permissions

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