Announcement

Collapse
No announcement yet.

The state of Radeon KMS on R600 hardware

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

  • The state of Radeon KMS on R600 hardware

    There is a working KMS code for R500 and older cards for now, but I can't find any to test my R600 hardware. Is there any code in a secret repo that enables KMS on R600? If not, when will it be possible to enable KMS on that hardware?

    Regards

  • #2
    Not yet AFAIK. The KMS code is working, but it needs memory management to be useful and the 6xx/7xx memory management is still WIP.
    Test signature

    Comment


    • #3
      @bridgman

      is it possible to give a timeframe of when things are going to be ready? (so far you have been accurate )


      also one more question

      how difficult will be the migration from normal mesa to gallium (for ATI cards)?


      Thanks in advance

      Comment


      • #4
        Best guess right now is 2.6.32. Don't think it will make 2.6.31, and we're going to try not to let it slip into 2.6.33.

        Until recently I didn't think we were going to need interrupts for the basic KMS / memory manager implementation but the devs are telling me that we do need them... which in turn means that we need to get another round of documentation and microcode out the door sooner than I had expected. Bah.

        All indications so far are that the transition from the current Mesa driver model to Mesa-over-Gallium3D should be fairly smooth, although it's too soon to be sure.

        The main reason we are still implementing on the "classic" driver model is that the Gallium3D code needs DRI2 and memory management, and we want to make sure that users on current distros (which, with the exception of Fedora, have neither DRI2 nor memory management for our parts) get 3D support as well.
        Last edited by bridgman; 04 June 2009, 03:49 PM.
        Test signature

        Comment


        • #5
          Originally posted by Wielkie G View Post
          There is a working KMS code for R500 and older cards for now, but I can't find any to test my R600 hardware. Is there any code in a secret repo that enables KMS on R600? If not, when will it be possible to enable KMS on that hardware?
          I doubt it'll be very secret when it comes out. With current interest levels I'd expect a parade all over Phoronix ATi section as soon as devs get anything worth testing done... We'll notice it when they're ready, don't worry.

          Comment


          • #6
            Originally posted by bridgman View Post
            Best guess right now is 2.6.32. Don't think it will make 2.6.31, and we're going to try not to let it slip into 2.6.33.

            Until recently I didn't think we were going to need interrupts for the basic KMS / memory manager implementation but the devs are telling me that we do need them... which in turn means that we need to get another round of documentation and microcode out the door sooner than I had expected. Bah.
            Now that you are here: I have a radeon hd3850 and I'm using a 2.6.30-rc8 kernel patched with drm-modesetting-radeon from Fedora 11. The kernel does the modesetting (it boots at 1680x1050 without any VESA driver in the kernel or vga=xxx boot parameter), but the driver ALSO seems to do the modesetting (the screen flickers when I startx or when I switch between X and vt). Is this caused by an incomplete kernel patch (i.e., the lack of "memory management" you mentioned, whatever that means :P) or because the xf86-video-radeon driver doesn't play nicely with KMS?

            Comment


            • #7
              You need a special version of the radeon driver as well; the code in master doesn't recognize KMS. If you're running KMS on a distro other than Fedora then Jerome's blog points to the latest bits you need :

              http://jglisse.livejournal.com
              Test signature

              Comment


              • #8
                Originally posted by bridgman View Post
                You need a special version of the radeon driver as well; the code in master doesn't recognize KMS. If you're running KMS on a distro other than Fedora then Jerome's blog points to the latest bits you need :

                http://jglisse.livejournal.com
                Thanks. I was already using the radeon-gem-cs3 branch (I changed the branch in the gentoo ebuild), but I was using r6xx-r7xx-support for libdrm instead of modesetting-gem (I am also using r6xx-r7xx-support for mesa, instead of radeon-rewrite, but I assume the mesa branch does not have influence on modesetting). Does the modesetting-gem branch of libdrm support r6xx?

                Comment


                • #9
                  Originally posted by Fran View Post
                  Thanks. I was already using the radeon-gem-cs3 branch (I changed the branch in the gentoo ebuild), but I was using r6xx-r7xx-support for libdrm instead of modesetting-gem (I am also using r6xx-r7xx-support for mesa, instead of radeon-rewrite, but I assume the mesa branch does not have influence on modesetting). Does the modesetting-gem branch of libdrm support r6xx?
                  libdrm is probably fine with r6xx, but there is no ttm drm code yet for r6xx/r7xx so I don't think the KMS enabled ddx will work at this point on r6xx/r7xx.

                  Comment


                  • #10
                    Originally posted by agd5f View Post
                    libdrm is probably fine with r6xx, but there is no ttm drm code yet for r6xx/r7xx so I don't think the KMS enabled ddx will work at this point on r6xx/r7xx.
                    Yep, I tried with the modesetting-gem branch of libdrm and it's the same as r6xx-r7xx-support. KMS works, but the driver simply does the modesetting again when starting X. Oh well, I can wait.

                    Comment

                    Working...
                    X