Announcement

Collapse
No announcement yet.

RadeonSI Gallium3D Gets UBO/TBO Support, OpenGL 3.3

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

  • RadeonSI Gallium3D Gets UBO/TBO Support, OpenGL 3.3

    Phoronix: RadeonSI Gallium3D Gets UBO/TBO Support, OpenGL 3.3

    Just ahead of the code branching of the Mesa 10.0 release this week, Marek Olk has committed a number of improvements to the open-source RadeonSI Gallium3D graphics driver for providing open-source to the Radeon HD 7000 series GPUs and newer on Linux...

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

  • #2
    Errata: there's no OpenGL 3.3 support for RadeonSI; there's no GLSL 1.50. RadeonSI has reached feature parity with r600g, however.

    Comment


    • #3
      Notably missing geometry shaders:

      Geometry shaders DONE (i965)

      That means GL 3.1.

      Comment


      • #4
        Something changed in the prettystacktrace function of the latest LLVM git and look like MESA can't use that function anymore.

        Commenting out the line #182 of src/gallium/auxiliary/gallivm/lp_bld_misc.cpp fix the compilation issue but a proper fix should be made...

        OpenGL renderer string: Gallium 0.4 on AMD TAHITI
        OpenGL core profile version string: 3.1 (Core Profile) Mesa 10.0.0-devel (git-4e0ed59)
        OpenGL core profile shading language version string: 1.40

        Comment


        • #5
          Thank you Marek! This was totally unexpected, I thought it would take months to reach feature parity with r600g. Thank you!!!! OpenGL 3.3 is near!!!

          Comment


          • #6
            Originally posted by RavFX View Post
            Something changed in the prettystacktrace function of the latest LLVM git and look like MESA can't use that function anymore.

            Commenting out the line #182 of src/gallium/auxiliary/gallivm/lp_bld_misc.cpp fix the compilation issue but a proper fix should be made...

            OpenGL renderer string: Gallium 0.4 on AMD TAHITI
            OpenGL core profile version string: 3.1 (Core Profile) Mesa 10.0.0-devel (git-4e0ed59)
            OpenGL core profile shading language version string: 1.40
            Yeah, there's a patch on the mesa-dev mailing list to fix that as of this morning... just hasn't been merged yet.

            Comment


            • #7
              Originally posted by Veerappan View Post
              Yeah, there's a patch on the mesa-dev mailing list to fix that as of this morning... just hasn't been merged yet.
              Good to know, was running into that very same issue, thanks.

              Comment


              • #8
                Originally posted by Alejandro Nova View Post
                RadeonSI has reached feature parity with r600g, however.
                Alright, I'm sold. As soon as all components have an official release, I'm ordering the new card.

                Thank you AMD!

                Comment


                • #9
                  Originally posted by wargames View Post
                  Thank you Marek! This was totally unexpected, I thought it would take months to reach feature parity with r600g. Thank you!!!! OpenGL 3.3 is near!!!
                  Well, HyperZ is still missing, but it's great that this landed in time for Mesa 10.

                  With some luck, 10.1 will have OpenGL 3.3 support for AMD across the board.

                  Comment


                  • #10
                    Originally posted by rohcQaH View Post
                    Alright, I'm sold. As soon as all components have an official release, I'm ordering the new card.

                    Thank you AMD!
                    Feature parity =/= efficiency parity. Everything could be looking correctly graphics wise while still running at 30% of the binary blob's speed. Don't get your hopes up yet.

                    Comment


                    • #11
                      Originally posted by Ancurio View Post
                      Feature parity =/= efficiency parity. Everything could be looking correctly graphics wise while still running at 30% of the binary blob's speed. Don't get your hopes up yet.
                      RadeonSI performance is pretty widely spread across the board. With Color Tiling enabled it can get up to 80% catalyst speeds with certain applications, though it averages more around 50-60%. Still not bad, all things considered.

                      Comment


                      • #12
                        Originally posted by dffx View Post
                        RadeonSI performance is pretty widely spread across the board. With Color Tiling enabled it can get up to 80% catalyst speeds with certain applications, though it averages more around 50-60%. Still not bad, all things considered.
                        If you are getting half performance of another software implementation of a driver, there are some fundamental missing features of the gpu you aren't using. Maybe if AMD would stop crippling their gpus with firmware blobs some of the Red Hat or Novell guys would help use the SI architecture to its potential, but as long as we only get face value FOSS drivers over obfuscated hardware we're stuck with the halfhearted efforts AMD is making.

                        Comment


                        • #13
                          Originally posted by zanny View Post
                          If you are getting half performance of another software implementation of a driver, there are some fundamental missing features of the gpu you aren't using. Maybe if AMD would stop crippling their gpus with firmware blobs some of the Red Hat or Novell guys would help use the SI architecture to its potential, but as long as we only get face value FOSS drivers over obfuscated hardware we're stuck with the halfhearted efforts AMD is making.
                          That's not true at all.

                          If you are running a game at 50fps instead of 100fps, it just shows that your driver has an extra 10ms of overhead per frame. Could that be because it's missing some hardware feature? Sure. It could also be because the driver is mistakenly pushing out an extra memory flush each frame. It could also be because of a bunch of cpu usage each frame. It could be because of a bunch of different issues.

                          The main hardware features that are currently missing are: HyperZ, proper-clocking (new kernel will enable it for some, but not all SI cards), and geometry shaders + GL4 features.

                          Could there be others? Sure - but one other thing you can be sure of is that the performance isn't simply due to AMD using firmware blobs. That's idiotic.

                          Comment


                          • #14
                            For whatever reason, the APUs perform amazing on r600g. it's starting to look like Radeon SI will be in shape for the Kaveri launch, which will get one more sale from me if the same thing applies.

                            Comment


                            • #15
                              Originally posted by zanny View Post
                              if AMD would stop crippling their gpus with firmware blobs some of the Red Hat or Novell guys would help
                              Novell's already been down that road with the radeonhd driver (and in case you don't remember, it didn't end well)...

                              Comment

                              Working...
                              X