Announcement

Collapse
No announcement yet.

Intel Is Using IGC In Their Windows Drivers, Internal Prototype For Mesa

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

  • Intel Is Using IGC In Their Windows Drivers, Internal Prototype For Mesa

    Phoronix: Intel Is Using IGC In Their Windows Drivers, Internal Prototype For Mesa

    At Intel's Architecture Day this week the company was talking about work on their new Windows graphics driver that is being timed for Xe but supporting existing generations of hardware as well. One of the interesting takeaways was seeing the Windows driver is now using the open-source "IGC" back-end...

    http://www.phoronix.com/scan.php?pag...dows-Mesa-Info

  • #2
    Link/version of this new (beta?) drivers package? Lastest (for Windows) is (beta) 8581 ( https://downloadcenter.intel.com/dow...?product=80939 )



    Comment


    • #3
      It's pretty awesome when Linux is on the leading edge of a graphics technology for a change. I'm looking forward to having both an advanced GPU and an open source driver stack.

      Comment


      • #4
        Benchmarks Benchmarks Benchmarks

        Seriously, I would like to see benchmarks when Mesa is coupled with IGC.

        Anything is better than LLVM if it can be done. Where's Valve?

        Comment


        • #5
          Originally posted by tildearrow View Post
          Benchmarks Benchmarks Benchmarks

          Seriously, I would like to see benchmarks when Mesa is coupled with IGC.

          Anything is better than LLVM if it can be done. Where's Valve?
          As stated in the article IGC is based on LLVM.

          It would be great to hear if they did some work on getting LLVM in better shape for GPUs though and if we see some of that work in upstream LLVM eventually.

          Comment


          • #6
            I think it would be pretty rad to see somebody take a serious look at getting accelerated Mesa running on Windows. At this point it's hard to argue that there is any transparency cost, since both AMD and Intel publish drivers well in advance of product launches, and with Vulkan, workloads are becoming more similar between the platforms.

            Comment


            • #7
              Sounds like a great idea to me. Why re-invent the wheel? Merge your assets and get a better result on all platforms.

              Comment


              • #8
                Originally posted by tildearrow View Post
                Anything is better than LLVM if it can be done.
                Why? You know there is a reason that a lot of people use LLVM. Just because some developers at Valve don't like it doesn't mean it's not a good compiler infrastructure.
                Last edited by agd5f; 08-13-2020, 06:19 PM. Reason: typo

                Comment


                • #9
                  [QUOTE=agd5f;n1200497]
                  Originally posted by tildearrow View Post
                  Anything is better than LLVM if it can be done./QUOTE]

                  Why? You know there is a reason that a lot of people use LLVM. Just because some developers at Valve don't like it doesn't mean it's not a good compiler infrastructure.
                  It's not a good online runtime compiler architecture, it's slow and horrible for runtime building for graphics shaders. This isn't just some developers at valve, there is lots of evidence in benchmarks.

                  Comment


                  • #10
                    rip my Haswell laptop :'(

                    Comment

                    Working...
                    X