GPU Offload in Rust: Portable, Safe, and Fast
Rust introduces a new framework for GPU programming that enhances memory safety while improving efficiency.
High-performance GPU programming has traditionally required a trade-off between execution efficiency and memory safety. Rust ensures compile-time memory safety for CPUs, but applying these constraints to GPU environments has often necessitated vendor-specific languages or unsafe pointers. This paper introduces a multi-vendor GPU compilation framework integrated into the Rust compiler (rustc) and LLVM backends.