Announcement

Collapse
No announcement yet.

Intel Iris Gallium3D Open-Source Driver Continues Speeding Along

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

  • Intel Iris Gallium3D Open-Source Driver Continues Speeding Along

    Phoronix: Intel Iris Gallium3D Open-Source Driver Continues Speeding Along

    Fresh out of the US holiday weekend, the Intel Iris Gallium3D driver that is forming as the company's future OpenGL Linux driver with better performance potential and modern design, saw a number of new code commits...

    http://www.phoronix.com/scan.php?pag...m3D-New-Monday

  • #2
    Have they wired up the vaapi or vdpau state trackers yet?

    Comment


    • #3
      Intels VAAPI driver is entirely independent of mesa, and I doubt that's going to change.

      Comment


      • #4
        Very exciting to see real progress on this. Is there any info on performance yet?

        Comment


        • #5
          Originally posted by BtbN View Post
          Intels VAAPI driver is entirely independent of mesa, and I doubt that's going to change.
          That's sad
          ## VGA ##
          AMD: X1950XTX, HD3870, HD5870
          Intel: GMA45, HD3000 (Core i5 2500K)

          Comment


          • #6
            Originally posted by darkbasic View Post

            That's sad
            Why? I don't see any benefit from adding an indirection layer via mesa there.

            Comment


            • #7
              Originally posted by BtbN View Post

              Why? I don't see any benefit from adding an indirection layer via mesa there.
              Its still preferable for general desktop stability to have shared codebases for things like hardware video. There have been plenty of times where (mostly on the AMD side) one of their implementations is non-conforming or broken and it is what lead a lot of software like OBS to take so long to enable it.

              A lot of that comes from duplicated work. Its why Intel would bother to invest so much effort in rewriting their driver for Gallium now, when several years ago when questioned about it Intel devs said it would have been too much effort to do.

              The longer you maintain a project for the most appreciative you are of shared code. When first starting out its a pain because its inflexible and may not conform to your mental model, but wait a few years and the technical debt of maintaining your own implementations will cost you a lot more in the long run.

              Comment


              • #8
                Originally posted by BtbN View Post
                Intels VAAPI driver is entirely independent of mesa, and I doubt that's going to change.
                This is something we've discussed actually, so I wouldn't be so sure. The existing libva implementation for Intel is not... great. Personally I'd love to have at least decode-only support wired up to the VA and VDPAU state trackers.

                Comment


                • #9
                  Arch users may install try it by mesa-intel-iris-git and lib32-mesa-intel-iris-git packages from AUR.

                  Comment


                  • #10
                    Originally posted by Venemo View Post
                    Very exciting to see real progress on this. Is there any info on performance yet?
                    Performance is still worse than i965, and that's expected - I haven't hooked up fast clears, color compression, hierarchical Z, or any of the other major bandwidth saving features. Been trying to finish up missing features and functional regressions first. That's getting closer, there are about 210 Piglit regressions (vs i965) and 560 OpenGL CTS failures. So, working on performance will probably come soon. I've also been cleaning up a bunch of prep work for merging, but there's more to do there too.
                    Free Software Developer .:. Mesa and Xorg
                    Opinions expressed in these forum posts are my own.

                    Comment

                    Working...
                    X