CGRA4HPC 2022 Invited Speaker: Mapping ML to the AMD/Xilinx AIE-ML architecture

Abstract

In the field of compute acceleration, machine learning model acceleration is one of the fastest growing areas of focus. ML model complexity in both compute and memory have driven the latest accelerator architectures and with that, developing ways to efficiently use these new architectures is the key to unlocking their potential. At AMD the AIE-ML architecture is our second generation AI-Engine architecture and we will dive into some of the ways we map the most important ML compute/bandwidth requirements to this architecture.