Announcement

Collapse
No announcement yet.

Valve's ACO Shader Compiler Under Review For The Mesa Radeon Vulkan Driver

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

  • #11
    Originally posted by timofonic View Post
    Please don't merge it until NIR improvements are made and can be used with OpenGL too.
    I get the NIR part. But why waiting before RadeonSI is ready to use ACO? This way it would postpone to enjoy the benefits of it until that work is ready. That would mean months wasted were you already could use ACO with Vulkan titles on RADV.

    Comment


    • #12
      Aco with Navi is still rough. If anyone wants to help testing, see: https://github.com/daniel-schuermann/mesa/issues/136

      Comment


      • #13
        Originally posted by timofonic View Post
        Please don't merge it until NIR improvements are made and can be used with OpenGL too.

        Any news about Intel and others doing an effort equivalent to ACO too? Please make it happen, LLVM sucks because of TOO MANY REASONS.
        Intel already has their own compiler which doesn't depend on LLVM and never did.

        Merging ACO as-is is a pretty significant milestone even if not all the optimizations are there yet.

        Comment


        • #14
          4k gaming here we come

          https://i.imgur.com/sJ4UnBK.jpg

          Comment


          • #15
            Originally posted by timofonic View Post
            Please don't merge it until NIR improvements are made and can be used with OpenGL too.

            Any news about Intel and others doing an effort equivalent to ACO too? Please make it happen, LLVM sucks because of TOO MANY REASONS.
            Yeah, we've very intentionally avoided LLVM in our software stack. It's never been required.
            Free Software Developer .:. Mesa and Xorg
            Opinions expressed in these forum posts are my own.

            Comment


            • #16
              Originally posted by brent View Post
              Yeah, the biggest advantage of ACO isn't performance of the compiled shaders or lower compilation time. It's breaking the dependency from LLVM! While Mesa will still use LLVM, it won't be as critical for AMDGPU as it is now. With ACO, we don't need to wait half a year for some compiler backend bugfix anymore.

              I'm still convinced that LLVM needs a severly revised release engineering process, though.
              yes right the biggest problem with my RX470 all the time was that if i want to upgrade the graphic driver the shit with the name LLVM needed to be updated to what makes a ton of problems.

              MESA with ACO instead of LLVM is really a big improvement.

              and i really do not care about the performance. bug-fixes and compatibility(what comes with newer driver) are much more important than the performance.
              Phantom circuit Sequence Reducer Dyslexia

              Comment


              • #17
                Originally posted by Qaridarium View Post

                yes right the biggest problem with my RX470 all the time was that if i want to upgrade the graphic driver the shit with the name LLVM needed to be updated to what makes a ton of problems.

                MESA with ACO instead of LLVM is really a big improvement.

                and i really do not care about the performance. bug-fixes and compatibility(what comes with newer driver) are much more important than the performance.
                I agree!

                But please don't forget about avoiding dependencies as much as possible (specially the biggest and more bug prone, such as LLVM) and improving in code reusability at same time.

                Any news about planned LLVM deprecation and obligatory usage of MESA's Gallium infrastructure? Please make it happen!!!

                Comment


                • #18
                  Originally posted by timofonic View Post

                  I agree!

                  But please don't forget about avoiding dependencies as much as possible (specially the biggest and more bug prone, such as LLVM) and improving in code reusability at same time.

                  Any news about planned LLVM deprecation and obligatory usage of MESA's Gallium infrastructure? Please make it happen!!!
                  A decentralized and distributed protocol, similar to Git and Kademlia.

                  Comment


                  • #19
                    Just did a quick test between Mesa-ACO and Mesa-Master with the ACO pull request...long story short is that those NIR optimizations in the Mesa-ACO repository cause serious pixelation and blur issues when aiming a weapon in Hitman 2 otherwise it works just fine

                    Also, Proton 4.11-5 doesn't have that crappy xscreensaver too many requests bug with amdgpu so my DS4's right stick is responsive and fully working again

                    I'd share a screenshot of it in effect but it's just a PITA to upload a screenshot to a random host and to then link to that here as well as the forum doesn't like showing my images when they're linked from Google Photos

                    EDIT: Whoops...I thought this was the other ACO thread

                    Comment


                    • #20
                      Originally posted by brent View Post
                      No, of course it does directly affect users. Bugfixes can be shipped much faster to users and with less bureaucracy.
                      no, that is indirect. users don't care which method is used to ship bugfixes fast. private llvm fork or just faster llvm releases in general would serve users just as well

                      Comment

                      Working...
                      X