The Peano AI engine fork is currently hosted via Xilinx/llvm-aie on GitHub, with full support for XDNA Phoenix / Hawk Point, while the XDNA2 support for the upcoming Ryzen AI 300 "Strix Point" hardware is still under development. AMD plans to upstream the relevant code into the LLVM codebase. The open-source compiler back-end for Ryzen AI NPUs is significant for making the accelerators more useful under Linux and for facilitating the upstreaming of their open-source XDNA driver into the Linux kernel.
Key takeaways:
- AMD announced a new project called Peano, an open-source LLVM compiler back-end for AMD/Xilinx AI engine processors, with a focus on the Ryzen AI SOCs.
- The Peano project will support the AIE2 architecture implemented by the XDNA accelerators in “Phoenix” and “Hawk Point” devices, with XDNA2 support for upcoming Ryzen AI 300 "Strix Point" hardware still being developed.
- AMD will be working to upstream the relevant code into the LLVM codebase, and the Peano AI engine fork is currently hosted via Xilinx/llvm-aie on GitHub.
- This open-source compiler back-end for Ryzen AI NPUs is significant in making the accelerators more useful under Linux and is also important for facilitating the upstreaming of their open-source XDNA driver into the Linux kernel.