MPM-LLM4DSE: Reaching the Pareto Frontier in HLS with Multimodal Learning and LLM-Driven Exploration

TL;DR

This paper introduces MPM-LLM4DSE, a novel framework combining multimodal learning and LLM-driven optimization to improve high-level synthesis design space exploration, achieving significant performance gains over existing methods.

Contribution

It proposes a multimodal prediction model and LLM-based optimizer with prompt engineering, advancing HLS DSE by better capturing semantic features and domain knowledge.

Findings

Outperforms state-of-the-art ProgSG by up to 10.25× in prediction accuracy.

Achieves an average of 39.90% performance gain in DSE tasks.

Demonstrates effectiveness of LLM-driven exploration with tailored prompts.

Abstract

High-Level Synthesis (HLS) design space exploration (DSE) seeks Pareto-optimal designs within expansive pragma configuration spaces. To accelerate HLS DSE, graph neural networks (GNNs) are commonly employed as surrogates for HLS tools to predict quality of results (QoR) metrics, while multi-objective optimization algorithms expedite the exploration. However, GNN-based prediction methods may not fully capture the rich semantic features inherent in behavioral descriptions, and conventional multi-objective optimization algorithms often do not explicitly account for the domain-specific knowledge regarding how pragma directives influence QoR. To address these limitations, this paper proposes the MPM-LLM4DSE framework, which incorporates a multimodal prediction model (MPM) that simultaneously fuses features from behavioral descriptions and control and data flow graphs. Furthermore, the framework employs a large language model (LLM) as an optimizer, accompanied by a tailored prompt engineering methodology. This methodology incorporates pragma impact analysis on QoR to guide the LLM in generating high-quality configurations (LLM4DSE). Experimental results demonstrate that our multimodal predictive model significantly outperforms state-of-the-art work ProgSG by up to 10.25$\times$. Furthermore, in DSE tasks, the proposed LLM4DSE achieves an average performance gain of 39.90\% over prior methods, validating the effectiveness of our prompting methodology. Code and models are available at https://github.com/wslcccc/MPM-LLM4DSE.