Offloading Rust To GPUs Proves Capable Of High Performance With Memory Safety

A research paper from Manuel S. Drehwald, Marcelo Domínguez, Kevin Sala, Alán Aspuru-Guzik, and Johannes Doerfert explores using LLVM's Offload infrastructure to run Rust code on GPU accelerators. Benchmarks show the safe Rust front-end, which avoids raw pointers in most kernels, performs competitively against hand-optimized CUDA and HIP C++ kernels, winning some tests and losing others. The researchers note further optimization opportunities in code generation and IR differences. Upstream tracking for Rust GPU offload support continues on a Rust-lang GitHub issue.

Source: https://www.phoronix.com/news/LLVM-Offload-Rust-Performance. 8 Sync News only summarizes and links out; content copyright belongs to the authors and original sources.