Rust SIMD on the GPU

VectorWare announced it can successfully use Rust's portable SIMD on GPUs, enabling developers to write high-performance applications using familiar Rust abstractions. The company realized that GPUs' SIMT (Single Instruction, Multiple Thread) model, where a warp's 32 lanes execute one instruction on different data, directly maps to SIMD's concept of one instruction operating on multiple data elements. This breakthrough allows developers to write SIMD code once using Rust's generic Simd type, which then compiles to the appropriate vector instructions whether targeting CPUs or GPUs.