AIE4ML: An End-to-End Framework for Compiling Neural Networks for the Next Generation of AMD AI Engines

TL;DR

AIE4ML is a comprehensive framework that automatically compiles neural networks into optimized firmware for AMD's next-generation AI engines, enabling high efficiency, scalability, and compatibility for ultra-low-latency AI inference.

Contribution

It introduces the first end-to-end compilation framework for AIE-ML devices, supporting multi-layer models, deterministic placement, and seamless integration with high-level tools.

Findings

Achieves up to 98.6% efficiency relative to single-kernel baseline.

Utilizes 97.4% of AIE tiles with entirely on-chip data movement.

Delivers GPU-class throughput with microsecond latency.

Abstract

Efficient AI inference on AMD's Versal AI Engine (AIE) is challenging due to tightly coupled VLIW execution, explicit datapaths, and local memory management. Prior work focused on first-generation AIE kernel optimizations, without tackling full neural network execution across the 2D array. In this work, we present AIE4ML, the first comprehensive framework for converting AI models automatically into optimized firmware targeting the AIE-ML generation devices, also with forward compatibility for the newer AIE-MLv2 architecture. At the single-kernel level, we attain performance close to the architectural peak. At the graph and system levels, we provide a structured parallelization method that can scale across the 2D AIE-ML fabric and exploit its dedicated memory tiles to stay entirely on-chip throughout the model execution. As a demonstration, we designed a generalized and highly efficient linear-layer implementation with intrinsic support for fused bias addition and ReLU activation. Also, as our framework necessitates the generation of multi-layer implementations, our approach systematically derives deterministic, compact, and topology-optimized placements tailored to the physical 2D grid of the device through a novel graph placement and search algorithm. Finally, the framework seamlessly accepts quantized models imported from high-level tools such as hls4ml or PyTorch while preserving bit-exactness. In layer scaling benchmarks, we achieve up to 98.6% efficiency relative to the single-kernel baseline, utilizing 296 of 304 AIE tiles (97.4%) of the device with entirely on-chip data movement. With evaluations across real-world model topologies, we demonstrate that AIE4ML delivers GPU-class throughput under microsecond latency constraints, making it a practical companion for ultra-low-latency environments such as trigger systems in particle physics experiments.