v1.0 - Initial release
Initial release of IRON and mlir-aie compiler
This project is primarily intended to support the open-source community with low-level access to AIE devices. We provide a Python-based programming flow: IRON supporting close-to-metal programming of AMD NPUs. This repository contains a programming guide and examples, however this project is not intended to represent an end-to-end compilation flow for all application designs. If you're looking for an out-of-the-box experience for highly efficient machine learning inference, check out AMD Ryzen™ AI Software.
🔧 Compiler and Toolchain Enhancements
- IRON Python API: Introduced a logical IRON API with support for deferred tile placement and data tiling helpers (Tensor Access Pattern Library).
- Just-In-Time (JIT) compilation: Added JIT compilation capabilities, enabling cleaner programming examples and simplified host code integration.
- Support for all current Ryzen™ AI NPUs: Phoenix, Hawk, Strix, Krackan Point.
📚 Documentation and Community Engagement
- Updated programming guides: The documentation has been updated to provide clearer documentation of IRON APIs including the logical deferred placement APIs.
- New mini tutorial: The programming guide now includes a mini tutorial to efficiently introduce IRON concepts with hands on exercises.
- Updated installation guides: The documentation has been updated to provide clearer instructions for installing the xdna-driver and setting up the development environment on various platforms.