- A next-generation MLIR-based compiler infrastructure.
- Aims to provide a portable and composable intermediate representation for GPU programming in ML and HPC workloads.
- A Python-based DSL built on MLIR
- Designed for writing high-performance GPU kernels with close-to-metal performance and integration with CUTLASS.
- Contributed to major compiler projects including:
- Clang and Flang, NVIDIA HPC Compilers (formerly PGI), MLIR, IREE
- Participated in the design of parallel programming models like: