MACO: A Multi-Agent LLM-Based Hardware/Software Co-Design Framework for CGRAs

2025-09-16

AI Summary

Coarse-grained reconfigurable arrays (CGRAs) suffer from an enormous design space, strong parameter coupling, and low efficiency in manual hardware/software co-optimization. Method: This paper proposes an automated hardware/software co-design framework for CGRAs based on multi-agent large language models (LLMs). The framework employs a分工-collaborative agent system integrating architecture-aware hardware/software partitioning, automated design error diagnosis and correction, PPA-driven candidate selection, and iterative feedback optimization. Contribution/Results: Its key innovation is a self-learning decision loop enabling end-to-end CGRA generation and joint optimization of performance, power, and area (PPA). Experimental evaluation on representative benchmarks shows that the generated CGRA architectures achieve superior PPA metrics compared to state-of-the-art LLM-based baselines and human-designed counterparts, while reducing design cycle time by over 60%. This significantly enhances both the efficiency and quality of custom accelerator design.

Technology Category

• Machine Learning: Hardware-aware ML
• Cognitive Modeling & Cognitive Systems: Agent Architectures
• Multiagent Systems: Coordination and Collaboration

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Economics, Online Markets and Human Computation: Architectures and workflows that use LLMs for crowd work
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Coarse-grained Reconfigurable Arrays (CGRAs) are a promising computing architecture that can deliver high-performance, energy-efficient acceleration across diverse domains. By supporting reconfiguration at the functional unit level, CGRAs efficiently adapt to varying computational patterns and optimize resource utilization. However, designing CGRAs is highly challenging due to the vast design space, independent architectural parameters, and the time-consuming nature of manual design. Fortunately, the rapid advancement of large language models (LLMs) presents new opportunities to automate this process. In this work, we propose MACO -- an open-source multi-agent LLM-based framework for Hardware/Software (HW/SW) co-design of CGRAs. The framework employs LLM reasoning to generate CGRAs across four stages: HW/SW co-design, Design error correction, Best design selection, and Evaluation & Feedback. Furthermore, MACO iteratively optimizes the generated CGRAs, leveraging agent reasoning and feedback to achieve higher PPA (that is, power, performance, and area) design points for a given domain. In addition, we introduce an LLM self-learning mechanism that employs LLM-driven decision making to select the optimal CGRA to accelerate the design process. We evaluate the framework with state-of-the-art LLM-based methods and manual CGRA design, in terms of performance, power consumption, and area. Experimental results show that MACO efficiently generates high-quality CGRA architectures, significantly reducing manual design effort and demonstrating the potential of our framework for real-world CGRA design.

Problem

• Automating CGRA design to overcome manual effort challenges
• Optimizing power, performance, and area through iterative LLM reasoning
• Selecting optimal CGRA architectures using multi-agent LLM framework

Innovation

• Multi-agent LLM framework for CGRA co-design
• Iterative optimization using agent reasoning and feedback
• LLM self-learning mechanism for optimal design selection