Sakhanda Wire
NVDA $153.59 -1.64% MSFT $507.00 -2.12% GOOGL $279.98 -1.32% META $619.94 -1.03% AMZN $256.62 -0.44%
← Back to the news

NVIDIA introduces two Rust pathways for CUDA GPU kernels

NVIDIA has announced CUDA Rust, an effort to make Rust a first-class language for writing GPU kernels. The initiative includes two open-source NVlabs projects covering CUDA’s programming models: cuda-oxide targets SIMT kernels, while cutile-rs targets the newer Tile model. Both compile Rust kernel code natively and use Rust ownership rules to detect buffer aliasing during compilation.

cuda-oxide is a custom rustc code-generation backend that transforms Rust MIR through Pliron and LLVM into PTX. It requires Linux, CUDA 12.x or newer, a GPU with compute capability 8.0 or later, clang with libclang, and the pinned nightly toolchain nightly-2026-04-03. Its SIMT approach provides explicit control over threads and memory, although shared memory currently requires unsafe code.

cutile-rs runs on stable Rust 1.89 or later and CUDA 13.3, with Linux and compute capability 8.0 or newer. It embeds kernel syntax in the host binary and JIT-compiles it through CUDA Tile IR at first launch. NVIDIA says the project is already used by Hugging Face’s Grout inference engine and mistral.rs. Both projects are in alpha and are not confirmed for production use.

Written by Sakhanda Wire AI from reporting by MarkTechPost. Facts, quotes and full context are in the original article below.

Originally published by MarkTechPost on

Read the original on MarkTechPost ↗

The reporting and the image belong to MarkTechPost and are credited and linked above; the text on this page is Sakhanda Wire's own summary.

← Back to the news

More stories

All the latest news