HLSmith: An Expert-Guided Agentic Framework for C/C++-to-HLS Translation

2026-08-07

AI Summary

This work addresses the limitations of current large language models, which lack hardware intuition and expert-level high-level synthesis (HLS) optimization knowledge, rendering them unreliable for translating C/C++ programs into high-performance, functionally correct hardware accelerators. The authors propose an expert-guided multi-agent framework that, for the first time, encodes established HLS development practices into an executable rule-based optimization library. By integrating multi-agent collaboration, feedback-driven orchestration, and tool-grounded model fine-tuning, the framework enables end-to-end automated translation and optimization. Furthermore, trajectory distillation is employed to convert optimization trajectories from commercial models into training data for open-source models. Evaluated on PolyBench, the approach achieves a 4.24× geometric mean speedup over ChatHLS, produces 100% functionally correct designs, and attains peak speedups of 252× with a commercial model and 138× with an open-source model.

Abstract

Application-specific FPGA accelerators offer substantial performance and energy-efficiency gains across many application domains, but developing them is costly, often requiring months of specialized effort. Even with high-level synthesis (HLS), designers still need extensive hardware expertise to build high-performance accelerators. Although large language models (LLMs) have demonstrated strong software-generation capabilities, even frontier models lack the hardware intuition and procedural knowledge needed to reliably translate baseline C/C++ programs into high-performance HLS designs: they struggle to identify effective architectures, follow the optimization processes used by HLS experts, and apply hardware transformations consistently across diverse kernels. We present HLSmith, an expert-guided framework for translating C/C++ programs into optimized HLS accelerators. HLSmith combines three components: an HLS optimization expertise library that encodes guarded transformation recipes, their applicability and prerequisite conditions, and unsafe cases to avoid; a staged, feedback-driven orchestration flow modeled on expert HLS development practice that guides agents through synthesis, bottleneck analysis, and optimization; and a tool-grounded model-adaptation pipeline that converts optimization trajectories from commercial frontier models into training data for fine-tuning open-weight LLMs. We evaluate HLSmith on PolyBench against ChatHLS, a leading prior agent-orchestration framework for HLS accelerator development. HLSmith achieves a geometric mean speedup of 4.24x over ChatHLS while producing functionally correct designs, in both software and RTL simulation, for every benchmark, compared with ChatHLS's 57% valid-design rate. It further reaches speedups of up to 252x and 138x with commercial frontier models and open-weight models, respectively.

Problem

• High-Level Synthesis
• FPGA Acceleration
• C/C++ to HLS Translation
• Hardware Design Automation
• Large Language Models

Innovation

• high-level synthesis
• LLM fine-tuning
• expert-guided optimization
• FPGA accelerator design
• code transformation