‹ BackHN Continuity

Thread

Nvidia announces native GPU programming in Rust

970 points · 404 comments · nonmaskable

  1. dllu · · focus · HN ↗
    Since NVIDIA owns huggingface now and huggingface has the excellent Candle [1] crate for inference on Rust, this seems like a good step towards nice native Rust kernels.

    [1] <a href="https:&#x2F;&#x2F;github.com&#x2F;huggingface&#x2F;candle" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;huggingface&#x2F;candle

    1. jacobgorm · · focus · HN ↗
      Nobody cares if kernels are written in Rust. Kernels were meant to be written in C, but if you want to go more high-level try Triton or a similar DSL that nicely abstract tile sizes etc.
      1. keithnz · · focus · HN ↗
        kernels aren&#x27;t meant to be written by any defined language. C is just a traditionally good default language that took over from assembly. No particular reason we have to stick with C.
        1. chadcmulligan · · focus · HN ↗
          And quite a few reasons that something better than C should be used. Rust seems a good candidate.
          1. jacobgorm · · focus · HN ↗
            What reasons would you have to prefer Rust over C for compute kernels? I am a great fan of Rust, but I don&#x27;t see any benefit for kernels, due to their relatively simple nature.
            1. chadcmulligan · · focus · HN ↗
              I&#x27;m not sure why being relatively simple would mean C over Rust? Rust still has the safety advantages.
Open on Hacker News to reply ↗

Unofficial Hacker News client; not affiliated with Y Combinator.