NVIDIA's CUDA platform for Rust. Host runtime crates plus Tile (cutile-rs) and SIMT (cuda-oxide) kernel programming models in idiomatic Rust.
LLM speculative inference server for heterogeneous hardware & consumer GPUs
Vx: one language, every chip.
NVIDIA's CUDA platform for Rust. Host runtime crates plus Tile (cutile-rs) and SIMT (cuda-oxide) kernel programming models in idiomatic Rust.
LLM speculative inference server for heterogeneous hardware & consumer GPUs
Vx: one language, every chip.