Lulzx/cuda-metal

CUDA on non-NVIDIA GPUs

Community maintained hardware plugin for vLLM on Apple Silicon

CUDA Tile IR is an MLIR-based intermediate representation and compiler infrastructure for CUDA kernel optimization, focusing on tile-based computation patterns and optimizations targeting NVIDIA tensor core units.

The CUDA target for Numba

Open-source CUDA, Triton and HIP compiler targeting multiple GPU and CPU architectures.

Metal programming in Julia

PMetal: high-performance Apple Silicon framework for local LLM inference, LoRA/QLoRA fine-tuning, serving, quantization, and MLX/Metal acceleration.

A JAX backend for Apple Metal Performance Shaders (MPS), enabling GPU-accelerated JAX computations on Apple Silicon.

Apple G13 GPU architecture docs and tools