LLM-DSE: Searching Accelerator Parameters with LLM Agents

Even though high-level synthesis (HLS) tools mitigate the challenges of programming domain-specific accelerators (DSAs) by raising the abstraction level, optimizing hardware directive parameters remains a significant hurdle. Existing heuristic and learning-based methods struggle with adaptability and sample efficiency. We present LLM-DSE, a multi-agent framework designed specifically for optimizing HLS directives. Combining LLM with design space exploration (DSE), our explorer coordinates four agents: Router, Specialists, Arbitrator, and Critic. These multi-agent components interact with various tools to accelerate the optimization process. LLM-DSE leverages essential domain knowledge to identify efficient parameter combinations while maintaining adaptability through verbal learning from online interactions. Evaluations on the HLSyn dataset demonstrate that LLM-DSE achieves substantial $2.55\times$ performance gains over state-of-the-art methods, uncovering novel designs while reducing runtime. Ablation studies validate the effectiveness and necessity of the proposed agent interactions. Our code is open-sourced here:this https URL.

@article{wang2025_2505.12188,
  title={ LLM-DSE: Searching Accelerator Parameters with LLM Agents },
  author={ Hanyu Wang and Xinrui Wu and Zijian Ding and Su Zheng and Chengyue Wang and Tony Nowatzki and Yizhou Sun and Jason Cong },
  journal={arXiv preprint arXiv:2505.12188},
  year={ 2025 }
}