A3C3: AI Algorithm and Accelerator Co-design, Co-search, and Co-generation

2026-06-18

AI Summary

This work addresses the longstanding disconnect between model development and hardware deployment in traditional AI systems, which often fails to balance accuracy, latency, energy efficiency, and resource utilization—particularly in heterogeneous, memory-intensive scenarios. To overcome this limitation, the authors propose a unified algorithm–accelerator co-design framework that enables end-to-end joint optimization across both neural network architectures and hardware implementation spaces for the first time. By integrating neural architecture search with hardware description language modeling through a differentiable or search-driven co-exploration mechanism, the framework provides a unified representation of software and hardware parameters. The resulting model–accelerator co-designed pairs consistently outperform conventional staged-design approaches across multiple performance metrics.

Abstract

We present a holistic methodology for artificial intelligence algorithm and accelerator co-design, co-search, and co-generation (A3C3), which jointly optimizes neural network architectures and their hardware implementations to address the inefficiencies of traditional top-down AI system design flows. Conventional AI deployment often treats model design and hardware mapping as separate stages: an algorithm is first developed for accuracy, and only afterward adapted to meet latency, throughput, energy, or resource constraints. This separation can lead to suboptimal systems, particularly as modern AI workloads become increasingly heterogeneous, memory-intensive, and platform-dependent. A3C3 instead parameterizes both algorithmic and accelerator design spaces and searches them jointly, enabling the automatic generation of model-accelerator pairs that better balance accuracy, latency, throughput, energy efficiency, and hardware utilization. This article is a book chapter of the Handbook of Embedded Machine Learning, edited by Sudeep Pasricha and Muhammad Shafique, Springer Nature.

Problem

• AI algorithm-accelerator co-design
• hardware-software co-optimization
• neural architecture search
• efficient AI deployment
• heterogeneous AI workloads

Innovation

• co-design
• co-search
• co-generation
• hardware-software co-optimization
• neural architecture search