NVIDIA Labs has released CUDA-Oxide 0.1, an experimental open-source Rust-to-CUDA compiler that allows developers to write SIMT GPU kernels in pure Rust without DSLs or foreign language bindings. The compiler takes standard Rust code and emits directly to NVIDIA's PTX format, using a custom rusc compiler back-end. Currently in alpha, it supports single-source compilation and device-side abstractions, though bugs and API breakage are expected as development continues.
Sort: