Evidence-Driven LLM Agent for C-to-Synthesizable-C Conversion and Verification

2026-06-24

AI Summary

This work addresses the challenge that general-purpose C programs often fail to pass through the four-stage high-level synthesis (HLS) pipeline due to violations of constraints imposed by the synthesizable C subset (HLS-C). To overcome this, the authors propose a closed-loop generate-validate-diagnose-repair framework that enables automatic transformation from C to HLS-C. The approach leverages multi-agent collaboration tightly integrated with a four-stage verifier and introduces three key innovations: an end-to-end agent workflow operating under evidence isolation, a progressive mismatch localization chain (PMLC) based on log normalization and AST backtrace slicing, and a typed two-stage evidence retrieval-augmented generation (RAG) mechanism powered by a self-evolving repair card pool. Experimental results demonstrate that the proposed method significantly outperforms state-of-the-art approaches in terms of repair success rate, robustness, and reproducibility.

Abstract

Software-compilable C programs routinely fail to complete the four-stage pipeline of a high-level synthesis (HLS) toolchain -- compilation, C simulation (CSim), synthesis, and C/RTL co-simulation (CoSim) -- because HLS accepts only a synthesizable subset of C (HLS-C). Yet most existing large language model (LLM) systems built for HLS code repair only cover the early pipeline stages and feed raw tool logs directly to the model, yielding brittle and hard-to-reproduce fixes. We formulate C-to-HLS-C conversion as a closed-loop generation-verification-diagnosis-repair problem on an HLS tool (Xilinx Vitis), contributing three components: an end-to-end workflow of cooperating agents closed by the four-stage verifier under strict evidence isolation; a Progressive Mismatch Localization Chain (PMLC) that localizes CSim/CoSim mismatches through log normalization, AST backward slicing, and dual-trace instrumentation; and a typed-query, two-stage evidence RAG backed by a self-evolving, family-routed repair-card pool. Experimental results show that the proposed workflow substantially outperforms all comparable state-of-the-art models.

Problem

• High-Level Synthesis
• Synthesizable C
• LLM Agent
• C-to-HLS Conversion
• Verification

Innovation

• Evidence-Driven LLM Agent
• C-to-HLS-C Conversion
• Progressive Mismatch Localization Chain
• Typed-Query RAG
• Closed-Loop Verification