Async/await on the GPU: Rust futures running on CUDA with VectorWare
Summary
VectorWare announces that Rust's Future trait and async/await can run on GPUs, arguing that futures provide a structured concurrency abstraction suitable for GPU execution. The article surveys competing approaches (JAX, Triton, CUDA Tile), discusses downsides of current models, and describes executors and future work to broaden GPU-native runtimes and multi-language support.