CubeCL: GPU Kernels in Rust for CUDA, ROCm, and WGPU

(github.com)

153 points | by ashvardanian 12 hours ago ago

24 comments

  • bionhoward 10 minutes ago

    Praying to the kernel gods for some Rust FP8 training

  • rfoo 2 hours ago

    I'd recommend having a "gemm with a twist" [0] example in the README.md instead of having an element-wise example. It's pretty hard to evaluate how helpful this is for AI otherwise.

    [0] For example, gemm but the lhs is in fp8 e4m3 and rhs is in bf16 and we want fp32 accumulation, output to bf16 after applying GELU.

    • ashvardanian an hour ago

      Agreed! I was looking through the summation example < https://github.com/tracel-ai/cubecl/blob/main/examples/sum_t...> and it seems like the primary focus is on the more traditional pre-2018 GPU programming without explicit warp-level operations, asynchrony, atomics, barriers, or countless tensor-core operations.

      The project feels very nice and it would be great to have more notes in the README on the excluded functionality to better scope its applicability in more advanced GPGPU scenarios.

      • 0x7cfe a minute ago

        CubeCL is the computation backend for Burn (https://burn.dev/) - ML framework done by the same team which does all the tensor magic like autodiff, op fusion and dynamic graphs.

  • gitroom 4 hours ago

    Gotta say, the constant dance between all these GPU frameworks kinda wears me out sometimes - always chasing that better build, you know?

  • kookamamie 3 hours ago

    This reminds me of Halide (https://halide-lang.org/).

    In Halide, the concept was great, yet the problems in kernel development were moved to the side of "scheduling", i.e. determining tiling/vectorization/parallellization for the kernel runs.

  • the__alchemist 9 hours ago

    Love it. I've been using cudarc lately; would love to try this since it looks like it can share data structures between host and device (?). I infer that this is a higher-level abstraction.

  • zekrioca 12 hours ago

    Very interesting project! I am wondering how it compare against OpenCL, which I think adopts the same fundamental idea (write once, run everywhere)? Is it about CUbeCL's internal optimization for Rust that happens at compile time?

    • fc417fc802 6 hours ago

      This appears to be single source which would make it similar to SYCL.

      Given that it can target WGPU I'm really wondering why OpenCL isn't included as a backend. One of my biggest complaints about GPGPU stuff is that so many of the solutions are GPU only, and often only target the vendor compute APIs (CUDA, ROCm) which have much narrower ecosystem support (versus an older core vulkan profile for example).

      It's desirable to be able to target CPU for compatibility, debugging, and also because it can be nice to have a single solution for parallelizing all your data heavy work. The latter reduces mental overhead and permits more code reuse.

      • zekrioca 4 hours ago

        Makes sense. And indeed, having OpenCL as a backend would be a very interesting extension.

    • nathanielsimard 10 hours ago

      A lot of things happen at compile time, but you can execute arbitrary code in your kernel that executes at compile time, similar to generics, but with more flexibility. It's very natural to branch on a comptime config to select an algorithm.

  • LegNeato 11 hours ago

    See also this overview for how it compares to other projects in the Rust and GPU ecosystem: https://rust-gpu.github.io/ecosystem/

  • adastra22 8 hours ago

    Where is the Metal love…

    • Almondsetat 6 hours ago

      Why would anyone love something born out of pure spite for industry standards?

      • m-schuetz an hour ago

        To be fair, the industry standards all suck except for CUDA.

      • pjmlp 5 hours ago

        For the same reason CUDA and ROCm are supported.

        • miohtama 2 hours ago

          Apple is known to be not that great contributor to open source, unlike Nvidia, AMD, Intel.

          • pjmlp an hour ago

            You should check Linus opinion on those.

            Also, to whom do you have to thank LLVM exists in first place, and has not fizzled out as yet another university compiler research project?

    • syl20bnr 8 hours ago

      It also compiles directly to MSL, it is just missing from the post title.

      • adastra22 6 hours ago

        No it compiles indirectly through wgpu, which means it doesn’t have access to any Metal extensions not exposed by the wgpu interface.

        • syl20bnr 17 minutes ago

          I am the coder of the MSL dialect for the CubeCL CPP compiler. Since 0.5 release it directly compiles to MSL and support simdgroup matrix functions for instance. It does use wgpu for the runtime but without naga as we added msl pass through to wgpu just for this.

    • moffkalast 7 hours ago

      From the moment I understood the weakness of my flesh, it disgusted me. I craved the strength and certainty of steel. I aspired to the purity of the Blessed Machine.