Announcement

Collapse
No announcement yet.

AMD R600/700 DRM Interrupts Support Pushed

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

  • AMD R600/700 DRM Interrupts Support Pushed

    Phoronix: AMD R600/700 DRM Interrupts Support Pushed

    One of the underlying features that has been lacking from the ATI R600/700 DRM / kernel mode-setting driver on Linux has been support for interrupts, which is needed for the sync-to-vblank operation and other important areas. Interrupts support has been lacking since AMD has not yet published any documentation concerning them for the Radeon HD 2000, 3000, and 4000 series ASICs, but today their code has cleared the legal review process for publishing...

    http://www.phoronix.com/vr.php?view=Nzc1MQ

  • #2
    Business as usual for our beloved opensource developers

    How much stuff that is basicly done is there currently waiting for the legal reviews process to finish?

    Comment


    • #3
      very nice. I hope the documentary will follow. And that means that I need to recompile my kernel from drm-next or any other branch?

      Comment


      • #4
        No vblank without interrups? I thought in Fedora 12 vblank is enabled by default and that causes a bit worse performance? How is that possible?

        I think I read somewhere that interrups were necessary for good power management. Hopefully powersaving becomes more effective.

        Thanks Alex for this nice code!

        Comment


        • #5
          Originally posted by bugmenot View Post
          No vblank without interrups? I thought in Fedora 12 vblank is enabled by default and that causes a bit worse performance? How is that possible?

          I think I read somewhere that interrups were necessary for good power management. Hopefully powersaving becomes more effective.

          Thanks Alex for this nice code!
          afaik interrupts are needed for changing the clock of the gpu memory, so you can change the clock inbetween rendering 2 frames...

          Comment


          • #6
            Originally posted by d2kx View Post
            Business as usual for our beloved opensource developers

            How much stuff that is basicly done is there currently waiting for the legal reviews process to finish?
            Not much -- just analog VGA support for Evergreen AFAIK. We tried to push all the important (and difficult) stuff through legal review earlier so that all the development work could be done in public repositories.

            It's not like there's a big heap of ready-to-go documentation just waiting to be released. All of the code and docs you see are being purpose-written for the open source effort, and until we get something working it's not clear what we need to review & release.

            Originally posted by bugmenot View Post
            No vblank without interrups? I thought in Fedora 12 vblank is enabled by default and that causes a bit worse performance? How is that possible?
            That's for 5xx/rs6xx and below, where interrupts were already supported. This new code does the same for 6xx/7xx.
            Last edited by bridgman; 11-30-2009, 04:14 PM.

            Comment


            • #7
              Hi everyone,

              great news for radeon oss, anyone know if it will be available on fedora 12 or we will need to recompile the kernel?
              Last edited by Sly154; 11-30-2009, 05:15 PM.

              Comment


              • #8
                Originally posted by bridgman View Post
                Not much -- just analog VGA support for Evergreen AFAIK.
                I'm afraid that would mean you missed HDMI fixes from Alex... :|

                Comment


                • #9
                  Whoops, yeah... make that "just analog VGA for Evergreen and HDMI audio patches for KMS"

                  Comment


                  • #10
                    I'll be happy the day I install Ubuntu with my ati hd 2600xt card, install foobillard and have 3d working (and without the ati bug that shows shadows around the table).

                    Nvidia: http://taufanlubis.files.wordpress.c...foobillard.png

                    Ati fglrx or open source driver: http://www.dedoimedo.com/images/game...obillard-3.jpg

                    Comment


                    • #11
                      I thought the conclusion from that discussion was that the game was using the wrong parameters (some kind of clamp mode ?), and that it was not a driver bug. Do you remember where the thread was ?

                      EDIT - found it : http://www.phoronix.com/forums/showt...t=17689&page=4
                      Last edited by bridgman; 11-30-2009, 05:34 PM.

                      Comment


                      • #12
                        Originally posted by bridgman View Post
                        I thought the conclusion from that discussion was that the game was using the wrong parameters (some kind of clamp mode ?), and that it was not a driver bug. Do you remember where the thread was ?

                        EDIT - found it : http://www.phoronix.com/forums/showt...t=17689&page=4
                        Yea you are right, I actually forgot the conclusion of that thread.... who knows if it will ever be fixed in official ubuntu rep..

                        Comment


                        • #13
                          This interrupts support also requires two new firmware files to be loaded by the open-source driver.
                          Can anyone please explain this to me?

                          The firmware is intended to be a patch to the kernel or is it a required bios update for tha graphic card? I really don't understand that sentence.
                          Thanks
                          Netrunner Linux - Rolling Release ; Nexus 5 ROM Chroma 5.1 ; NAS 6TB on FreeNAS

                          Comment


                          • #14
                            None of the above, actually

                            ATI graphics chips use writable microcode stores for the hardware state machines (rather than permanently burning the microcode into the chip) so microcode images need to be loaded by the driver at startup. Two microcode images are already being loaded for 6xx/7xx, one for the "cp" block and another for the "pfp" block.

                            Supporting interrupts requires that a third microcoded block ("rlc") be enabled, which requires both driver code and an additional microcode image. There are two new files -- one for 6xx hardware and one for 7xx -- and the driver loads whichever one is appropriate for the hardware it finds.

                            These files aren't really "firmware" in the traditional sense, but there's a lack of agreement on what exactly they *should* be called
                            Last edited by bridgman; 11-30-2009, 05:47 PM.

                            Comment


                            • #15
                              Oh, I see
                              Thank you John.
                              Very interesting
                              Netrunner Linux - Rolling Release ; Nexus 5 ROM Chroma 5.1 ; NAS 6TB on FreeNAS

                              Comment

                              Working...
                              X