Announcement

Collapse
No announcement yet.

4770HD: Various bugs and speed issues with KMS

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

  • 4770HD: Various bugs and speed issues with KMS

    Hi,

    I would like to know if I'm the only one to encounter some problems I run into when a 2.6.31 kernel with drm-linus (as explained somewhere on freedesktop).
    Here is the list of the problems I run into:

    First of all, suspend doesn't work. I was expecting it to work as intended when David Airlie committed a few patches but it didn't :s.

    Games stopped working using KMS. Dmesg gives me this :
    "[drm:r600_cs_packet_parse] *ERROR* Can not parse packet at 597 after CS end 597 !
    [drm:r600_packet3_check] *ERROR* bad DRAW_INDEX
    [drm:radeon_cs_ioctl] *ERROR* Invalid command stream !"
    -------
    Edit: This has been fixed when recompiling everything once again from scratch. The problem still remains in a non-kms environment.
    -------

    Also, after a few hours of use, I start getting screen corruption like this: http://fs.mupuf.org/posted_images/r7...corruption.png

    The last problem I run into is the time gtkperf takes to perform.
    GtkPerf 0.40 - Starting testing: Thu Oct 15 17:34:35 2009

    GtkEntry - time: 0,13
    GtkComboBox - time: 4,17
    GtkComboBoxEntry - time: 2,79
    GtkSpinButton - time: 0,93
    GtkProgressBar - time: 0,72
    GtkToggleButton - time: 1,27
    GtkCheckButton - time: 0,61
    GtkRadioButton - time: 0,85
    GtkTextView - Add text - time: 0,39
    GtkTextView - Scroll - time: 0,98
    GtkDrawingArea - Lines - time: 4,34
    GtkDrawingArea - Circles - time: 11,22
    GtkDrawingArea - Text - time: 40,88
    GtkDrawingArea - Pixbufs - time: 3,25
    ---
    Total time: 72,54

    This is something that used to be quite good (It took 13 seconds to perform this test on Sept. 25, 2009) but I don't really know when It broke. I've been quite busy and so, I didn't took the time to bisect the regression.

    Am I the only one ? Could this be because of my packaging ?

    Thanks in advance
    Last edited by MPF; 10-16-2009, 11:48 AM.

  • #2
    Originally posted by MPF View Post
    Games stopped working using KMS. Dmesg gives me this :
    "[drm:r600_cs_packet_parse] *ERROR* Can not parse packet at 597 after CS end 597 !
    [drm:r600_packet3_check] *ERROR* bad DRAW_INDEX
    [drm:radeon_cs_ioctl] *ERROR* Invalid command stream !"
    I think using/updating git Mesa should fix that one.

    Comment


    • #3
      Originally posted by legume View Post
      I think using/updating git Mesa should fix that one.
      FWIW, I filed a bug for Fedora Rawhide/12 about some of these issues. No, they're not using the very latest bits from git, but it really sucks that users with common hardware will have a horrifically broken experience out of the box with Fedora as things stand right now.

      https://bugzilla.redhat.com/show_bug.cgi?id=524419

      The desktop can last a long while without corruption, but getting into video playback, lots of Firefox tabs, Flash applets in Firefox, or anything else that seems to stress video memory allocation ends up resulting in a ton of corrupted video memory buffers. Kind of annoying to say the least.

      Comment


      • #4
        Originally posted by legume View Post
        I think using/updating git Mesa should fix that one.
        That's the problem, it doesn't. I recompile the graphic stack almost every day (thanks to a wonderful script).

        Originally posted by elanthis View Post
        FWIW, I filed a bug for Fedora Rawhide/12 about some of these issues. No, they're not using the very latest bits from git, but it really sucks that users with common hardware will have a horrifically broken experience out of the box with Fedora as things stand right now.

        https://bugzilla.redhat.com/show_bug.cgi?id=524419

        The desktop can last a long while without corruption, but getting into video playback, lots of Firefox tabs, Flash applets in Firefox, or anything else that seems to stress video memory allocation ends up resulting in a ton of corrupted video memory buffers. Kind of annoying to say the least.
        Nice to see that the problem is reported.
        There is something strange, even if gtkperf is slow as hell, my KDE desktop is quite snappy, maybe an update slowed a code-path GTK use heavily.

        What about suspend, does it work on your computer ?

        Comment


        • #5
          Hi,

          I'm interested in your wonderful script to compil graphic stack from git.
          Could you post it ?

          Bibi

          Comment


          • #6
            Originally posted by bibi View Post
            Hi,

            I'm interested in your wonderful script to compil graphic stack from git.
            Could you post it ?

            Bibi
            It is ArchLinux related, are you an archer ?

            Comment


            • #7
              Update:

              3D/compositing doesn't work anymore when using a non-KMS kernel.

              Any idea ?

              Edit:
              Some more information about this problem:

              $ LIBGL_DEBUG=verbose glxgears
              libGL: XF86DRIGetClientDriverName: 4.3.0 r600 (screen 0)
              libGL: OpenDriver: trying /usr/lib/xorg/modules/dri/tls/r600_dri.so
              libGL: OpenDriver: trying /usr/lib/xorg/modules/dri/r600_dri.so
              drmOpenDevice: node name is /dev/dri/card0
              drmOpenDevice: open result is 4, (OK)
              drmOpenByBusid: Searching for BusID pci:0000:02:00.0
              drmOpenDevice: node name is /dev/dri/card0
              drmOpenDevice: open result is 4, (OK)
              drmOpenByBusid: drmOpenMinor returns 4
              drmOpenByBusid: drmGetBusid reports pci:0000:02:00.0
              libGL: Can't open configuration file /etc/drirc: No such file or directory.
              libGL: Can't open configuration file /home/mupuf/.drirc: No such file or directory.
              drmRadeonCmdBuffer: -22. Kernel failed to parse or rejected command stream. See dmesg for more info.

              dmesg :
              ...
              [drm:radeon_cp_indirect] *ERROR* sending pending buffer 4
              [drm:radeon_cp_indirect] *ERROR* sending pending buffer 30
              [drm:radeon_cp_indirect] *ERROR* sending pending buffer 1
              [drm:radeon_cp_indirect] *ERROR* sending pending buffer 31
              [drm:radeon_cp_indirect] *ERROR* sending pending buffer 23
              [drm:radeon_cp_indirect] *ERROR* sending pending buffer 26
              [drm:radeon_cp_indirect] *ERROR* sending pending buffer 12
              [drm:radeon_cp_indirect] *ERROR* sending pending buffer 12
              [drm:radeon_cp_indirect] *ERROR* sending pending buffer 24
              [drm:radeon_cp_indirect] *ERROR* sending pending buffer 3
              [drm:radeon_cp_indirect] *ERROR* sending pending buffer 22

              Xorg.0.log --> nothing to see

              Hope this will help !!
              Last edited by MPF; 10-16-2009, 11:45 AM.

              Comment


              • #8
                I also forgot to mention that the option EXANoDownloadFromScreen is still needed if you don't want to see screen corruption all over the screen and pixmaps.

                Comment

                Working...
                X