NVIDIA’s CUDA Rust effort lets developers write GPU kernels in Rust instead of switching to another language. It offers a low-level SIMT path through cuda-oxide and a higher-level Tile path through cutile-rs, bringing Rust’s ownership model closer to GPU execution.
A batched compact-Householder QR kernel is tricky on GPUs because Householder reflectors form a dependency chain. This post explains how blocked Householder QR with compact WY updates turns the trailing work into GEMM-shaped operations, and how an auto-research loop with Codex drove fast, correct kernel iterations to a 232× speedup over baseline.