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. - NVIDIA/cuda-tile

Hacker News is a community-driven platform for sharing and discussing technology news, startups, and programming-related topics. Through user submissions and comments, Hacker News offers insights into emerging technology trends, industry developments, and entrepreneurial ventures. Readers can participate in discussions, share their insights, and stay informed about the latest advancements in technology and innovation.

Hacker News

NVIDIA released CUDA Tile IR, an open-source MLIR-based compiler infrastructure for optimizing CUDA kernels through tile-based computation patterns targeting tensor cores. The project includes a domain-specific MLIR dialect, Python bindings, bytecode serialization, and comprehensive testing. It requires CMake 3.20+, C++17, and compatible MLIR/LLVM sources. The release aligns with CUDA Toolkit 13.1 and provides tools for programmatic IR construction, transformation, and integration into existing projects via pre-built libraries or source integration.

NVIDIA/cuda-tile: CUDA Tile IR is an MLIR-based intermediate representation and compiler infrastructure for CUDA kernel optimization, focusing on tile-based computation patterns and optimizations targ

Example: Writing and Running a CUDA Tile IR Program