Announcement

Collapse
No announcement yet.

ZLUDA Takes On Third Life: Open-Source Multi-GPU CUDA Implementation Focused On AI

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

  • ZLUDA Takes On Third Life: Open-Source Multi-GPU CUDA Implementation Focused On AI

    Phoronix: ZLUDA Takes On Third Life: Open-Source Multi-GPU CUDA Implementation Focused On AI

    The open-source ZLUDA project began life as a drop-in CUDA replacement that ran atop Intel GPUs using the Level Zero API. Then AMD quietly began funding it for several years as a viable CUDA implementation running atop AMD GPUs until discontinued funding earlier this year. ZLUDA for AMD GPUs was then made open-source but then in August the ZLUDA code was removed at AMD's request. Today it's taking on its third incarnation...

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    With Nintendo destroying their emulators left and right, I don't see a bright future for this project unless NVIDIA officially confirms that they're not opposed to its existence.

    AFAIK they have changed the CUDA licence agreement to not allow such projects. Anyways, as long as the principle developer is being paid for this work, I guess it's ... fine? As long as NVIDIA doesn't sue the living crap out of him.

    Comment


    • #3
      It may be that parts of CUDA license agreement that add additional restrictions are not enforceable in EU where Andrzej is living and developing ZLUDA. He must have looked into it before deciding to make it his business.
      Last edited by reavertm; 04 October 2024, 11:56 AM.

      Comment


      • #4
        I don't really see how CUDA's license agreement is relevant regardless of where he's from. If you haven't installed CUDA and/or do not agree to the license and you are not using any of Nvidia's CUDA code then what legal power do they even have? Unlike a Nintendo emulator, the primary use of ZLUDA isn't obviously used for piracy, and since it isn't [yet] targeting Nvidia GPUs then at that point, it has practically nothing to do with Nvidia at all.

        Think of it like the Mono project vs .NET. MS would have an incentive to shut that down, but they didn't, probably because they had no power to do so.

        Comment


        • #5
          Originally posted by reavertm View Post
          It may be that parts of CUDA license agreement that add additional restrictions are not enforceable in EU where Andrzej is living and developing ZLUDA. He must have looked into it before deciding to make it his business.
          Yes. It is lawful in France, for instance, to reverse engineer a product you own (with some restrictions).
          It is also lawful to reimplement APIs.
          Finally, algorithms can't be copyrighted.

          Comment


          • #6
            Originally posted by avis View Post
            With Nintendo destroying their emulators left and right, I don't see a bright future for this project unless NVIDIA officially confirms that they're not opposed to its existence.

            AFAIK they have changed the CUDA licence agreement to not allow such projects. Anyways, as long as the principle developer is being paid for this work, I guess it's ... fine? As long as NVIDIA doesn't sue the living crap out of him.
            He would have never agreed to such terms if he had never installed CUDA software. He is just writing code to compile code to OneAPI/Rocm

            Comment


            • #7
              Originally posted by brunosalezze View Post

              He would have never agreed to such terms if he had never installed CUDA software. He is just writing code to compile code to OneAPI/Rocm
              AFAIK ZLUDA allows to run unmodified CUDA binaries on AMD GPUs. That's quite different from "recompilation".

              Comment


              • #8
                Originally posted by brunosalezze View Post

                He would have never agreed to such terms if he had never installed CUDA software. He is just writing code to compile code to OneAPI/Rocm
                Well, he is actually doing 2 things:

                1. Implementing CUDA standard APIs on other GPUs, just like Google implemented the Java standard libraries on their own JVM in Android, Google got sues By Sun but won the case because it was decided that interfaces are not covered by copyright, only implementation.

                2. Creating new compiler backends for CUDA. Here is how CUDA works:

                . CUDA source code in compiled into NVIDIA PTX, a intermediate representation, similar to LLVM IR or Java bytecode, it is done for portability, as GPUs generally change their ISA every 1 or 2 generations, not being backwards compatible like x86.

                . NVIDIA PTX code is compiled to machine code appropiate for the users' GPU, it is compiled on the user's side. (A compiler bundled with the drivers, that is how OpenGL, Vulkan, DirectX and any GPU API works).

                So the developer is just creating a new compiler backend, just like the Mono project, wich made C# able to run on Linux by compiling CIL. Never got sued by Microsoft AFAIK.

                Comment


                • #9
                  Originally posted by avis View Post
                  AFAIK ZLUDA allows to run unmodified CUDA binaries on AMD GPUs. That's quite different from "recompilation".
                  Yeah, sounds like Wine for Windows software, which exists.

                  Comment


                  • #10
                    Too bad AMD didn't actually challenge this. The government is already going after Nvidia as a monopoly and I doubt they can even attack a competitor right now.

                    Comment

                    Working...
                    X