Announcement

Collapse
No announcement yet.

The Grinch That Stole The Radeon Gallium3D Performance?

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • The Grinch That Stole The Radeon Gallium3D Performance?

    Phoronix: The Grinch That Stole The Radeon Gallium3D Performance?

    There are some significant performance drops right now on Mesa master for the forthcoming 7.12/8.0 release concerning the Gallium3D driver for older ATI Radeon graphics processors. The performance of the R300g driver is now setback compared to earlier Mesa releases.

    http://www.phoronix.com/vr.php?view=16861

  • HokTar
    replied
    Originally posted by FireBurn View Post
    Sigh

    Phoronix - where articles ARE the bug reports
    Phoronix - where there _are_ articles about r300g.

    Leave a comment:


  • Kayden
    replied
    Just export LIBGL_DRIVERS_PATH=/path/to/mesa-source/lib and then export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$LIBGL_DRIVERS_PA TH. There's no need for LD_PRELOAD. Then, you don't have to 'make install'.

    Leave a comment:


  • bridgman
    replied
    Originally posted by zwastik View Post
    Gallium 3d developers have implemented more and more features to their drivers and that is great, but sometimes I think that this has been at the cost of performance (which is far behind proprietary drivers).
    I don't think the developers knowingly implement anything which hurts performance, except in cases where frequently used cases are sped up at the expense of a less common case which gets slowed down.

    It's more likely that the developers implemented a new feature, application runtime logic saw the new feature and used it, but for that particular application/card/driver combination using the new feature is actually slower than running without the feature. Game developers do often put heuristics in which help to choose the right code paths depending on GPU/CPU/driver but they tend to be backward looking and have trouble dealing with a driver that is constantly evolving.

    Originally posted by zwastik View Post
    Also, I am curious of how much attention do older hardware drivers get. Maybe this is not an issue for some, but here in third world countries we run a lot of older hw in our machines.
    Since open source developers often start as users who "get fed up and decide to take matters into their own hands" I imagine you will see more developers appearing in those areas as well. Open source projects tend to be "self correcting" that way, although the high learning curve for GPU drivers does make it harder for new developers to step in as existing developers move onto other projects.

    Leave a comment:


  • Qaridarium
    replied
    yes great work ! you are a hero

    Leave a comment:


  • DaemonFC
    replied
    Originally posted by Michael View Post
    The regressed commits should be published on Christmas.
    Don't you actually mean this is an R300g issue, which does not include support for any RadeonHD models?

    I notice you didn't check out r600g, which does support all the RadeonHD models.

    Also worth a mention but not really a big issue is r100 and r200, but those cards are so old that I doubt anyone using them is running anything too stressful on the GPU.

    Leave a comment:


  • TheCycoONE
    replied
    Originally posted by FireBurn View Post
    Sigh

    Phoronix - where articles ARE the bug reports
    Bug reports that generate revenue maintain the tester's income and interest, as well as awareness in the community. In the graphics community a lot of the developers read Phoronix so the message isn't being lost (as evidence, it looks like they already found this regression).

    Kudos Michael.

    Leave a comment:


  • del_diablo
    replied
    Originally posted by FireBurn View Post
    Sigh

    Phoronix - where articles ARE the bug reports
    Phoronix, where they post better articles than just reprints of ads.

    Leave a comment:


  • Michael
    replied
    Originally posted by zwastik View Post
    This analysis confirms that performance is an important issue the gallium3d drivers need to address. Gallium 3d developers have implemented more and more features to their drivers and that is great, but sometimes I think that this has been at the cost of performance (which is far behind proprietary drivers).
    Also, I am curious of how much attention do older hardware drivers get. Maybe this is not an issue for some, but here in third world countries we run a lot of older hw in our machines.
    Anyway I am confident that these performance issue will be resolved. =)

    Merry Christmas.
    https://twitter.com/#!/michaellarabe...60220548927488
    https://twitter.com/#!/michaellarabe...62664813146112

    Leave a comment:


  • zwastik
    replied
    This analysis confirms that performance is an important issue the gallium3d drivers need to address. Gallium 3d developers have implemented more and more features to their drivers and that is great, but sometimes I think that this has been at the cost of performance (which is far behind proprietary drivers).
    Also, I am curious of how much attention do older hardware drivers get. Maybe this is not an issue for some, but here in third world countries we run a lot of older hw in our machines.
    Anyway I am confident that these performance issue will be resolved. =)



    Merry Christmas.

    Leave a comment:

Working...
X