#gpuprogramming

Nvidia announced two projects — cuda-oxide and cutile-rs — for writing GPU kernels in Rust compiled directly to PTX. I go through what the primary source says, what's still untested, and why "native" isn't the same as "replacement for CUDA C++."

Modular put Mojo under Apache 2.0 with LLVM exceptions on August 18, 2026. Here is what the release changes for local AI builders, and what it does not.

Rust GPU offload now works without unsafe code. Real benchmarks: 11% faster to 46% slower than CUDA on an H100, and a transfer bug that costs 400x more.

The CPU optimizes for Latency through complex Branch Prediction and Out-of-Order execution; the GPU...