AMD Versal AI-Engines for fixed latency environments

TL;DR

This paper evaluates AMD Xilinx's Versal AI-Engine co-processors for fixed latency data processing in high-energy physics, demonstrating their suitability for ML tasks like BDT and CNN in real-time environments.

Contribution

It provides a technical assessment of AIE deployment in fixed latency settings, showcasing their potential as alternatives to traditional FPGA-based solutions for ML applications.

Findings

AIEs can meet fixed latency requirements in high-energy physics experiments.

Vectorized BDT and CNN implementations perform effectively on AIEs.

AIEs offer a feasible solution for real-time ML processing in sensor proximity environments.

Abstract

Complex, high-throughput data acquisition and processing systems, such as those used in high-energy physics experiments, are increasingly moving sophisticated pattern recognition and data compression algorithms closer to the sensors themselves. To meet these needs, programmable device manufacturers offer multi-silicon die packages that commonly include dedicated co-processors within the same package. We present a technical study of a new family of such co-processors from AMD Xilinx, the Adaptive Intelligence (AI) Engine, or AIE, as part of the Versal architecture. Specifically, we focus on the deployment capabilities of AIEs in fixed latency environments such as those typically found in colliding beam experiments like those at the Large Hadron Collider. We evaluate the performance of a vectorised implementation of both a Boosted Decision Tree (BDT) and a Convolutional Neural Network (CNN), thereby demonstrating the feasibility of deploying AIEs for ML applications in such environments and their use as possible alternatives to traditional programmable logic-based implementations.