>CUDA-oxide: Nvidia's official Rust to CUDA compiler>https://nvlabs.github.io/cuda-oxide/index.html>cuda-oxide is an experimental Rust-to-CUDA compiler that lets you write (SIMT) GPU kernels in safe(ish), idiomatic Rust. It compiles standard Rust code directly to PTX — no DSLs, no foreign language bindings, just Rust.>coda-oxide sets up the future of computation done right, in rust.
CUDA is basically C beause on GPUs you cannot do the same fuckeries as in sepples. Since Rust is a substitute of C and not C++, it makes sense to expand in that direction
Rust is the premier LLM code generation language, so it makes sense.
>>108801568Why do they push CUDA so much for this? All it has is CUDA calling conventions for the generated kernels.With this Rust replaces CUDA, not compiles to it.
>>108801568Cool shit but i'd much rather like a cuda like api that works over vulkan so it's gpu agnostic.
I've never seen another language be around for so long, become so popular, but still have such a massive inferiority complex that manifests in a constant need for approval and validation by its detractors.
