Reproduce “A2H-MAS: An Algorithm-to-HLS Multi-Agent System for Automated and Reliable FPGA Implementation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Jie Lei, Ruofan Jia, J. Andrew Zhang, Hao Zhang
A2H-MAS is presented as an automated, multi-agent system that reliably translates MATLAB algorithms into high-quality HLS C++ for FPGA implementation, prioritizing hardware efficiency and performance. The system demonstrated its effectiveness on 5G NR SSB detection and WLAN synchronization, producing designs with efficient resource usage and high operating frequencies, and achieving significant latency and resource reductions compared to direct translation.
Key insights
- Strategic algorithmic transformations (e.g., streaming architectures, specialized arithmetic structures) contribute more significantly to hardware efficiency than solely applying HLS pragmas.
- A multi-agent system with modular design, standardized interfaces, and deterministic, tool-driven verification effectively mitigates LLM limitations like hallucinations and unreliability in complex engineering workflows.
- Decomposing complex algorithms into dataflow-oriented submodules simplifies the conversion process, enables targeted optimization, and improves the scalability and maintainability of the generated hardware.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.