NVIDIA launches CUDA Rust: two tracks for writing GPU kernels in Rust
blelbach · x · 2026-09-09
NVIDIA announced CUDA Rust with two tracks: cuda-oxide, a custom rustc codegen backend compiling SIMT-style kernels directly to PTX via Pliron IR and LLVM (early alpha, pinned nightly toolchain), and cutile-rs, tile-based GPU programming on stable Rust 1.89+ with CUDA 13.3, where the compiler handles thread mapping and memory layout via CUDA Tile IR JIT. Already on crates.io and used in HuggingFace's Grout inference engine and mistral.rs. Both enforce compile-time memory safety (DisjointSlice/launch contracts vs tensor partitioning). NVIDIA plans interoperability with CUDA C++ and Python.
Related event: NVIDIA Launches CUDA Rust with Two Paths for Writing GPU Kernels(5 posts)→
More from Infra
- Open-source Latch compiles PyTorch into training megakernels, 2.75x faster on H100 — sguada · 2026-10-27
- NVIDIA Claims HGX Rubin NVL8 Delivers 10x More Tokens per Megawatt Than B200 — thursdai_pod · 2026-10-27
- DeepSpeed shows AutoEP/AutoTP/AutoSP parallelism gains, hosts office hours — PyTorch · 2026-10-27
- Running a local 9B as a coding agent: brief the big model to plan, let the small model execute — Electrical_Ninja3805 · 2026-10-27
- Cloudflare Durable Objects validated as top architecture for AI agent WebSocket servers — tristanbob · 2026-10-27
- Gary Marcus: With Q3 data added, Anthropic's compute-efficiency growth shows deceleration — GaryMarcus · 2026-10-27