LACE: Large Language Model Aided Multi-Agent Framework for Agile RISC-V Instruction Extension

2026-08-03

AI Summary

This work addresses the challenges of adapting RISC-V instruction set extensions in multicore implementations, where manual integration is cumbersome and differential testing is fragile. The authors propose the first large language model–driven multi-agent automation framework that translates natural language extension specifications into a two-tier intermediate representation—operational-level and HDL task-level. By combining retrieval-augmented local RTL editing with compiler-agnostic formal verification (leveraging riscv-formal/RVFI), the framework enables end-to-end verifiable RTL generation. The approach supports cross-core reuse and rapid integration, achieving a pass@1 accuracy of 72.8% across four embedded RISC-V cores—substantially improving over near-zero baselines—while enhancing code localization precision and reducing rework.

Abstract

Domain-specific Instruction Set Architecture eXtensions (ISAX) are widely adopted in the RISC-V ecosystem to accelerate emerging workloads, but implementing and validating ISAXes across different cores remains slow and fragmented. Existing frameworks still require per-core interface adaptation, and differential testing often breaks once either the microarchitecture or the ISAX changes. We present LACE, an LLM-aided multi-agent workflow that translates natural-language ISAX intents into a compact two-level IR (operation-level and HDL task-level), performs retrieval-guided localized RTL edits over large repositories, and closes the loop with a compiler-agnostic riscv-formal checking flow (assuming RVFI availability or instrumentation). Across four embedded RISC-V cores, LACE raises pass@1 generation accuracy from near-zero to 72.8\% within our evaluation setup, while improving code localization and reducing integration rework. The code of LACE is available at https://github.com/UMN-ZhaoLab/LACE.

Problem

• RISC-V
• Instruction Set Architecture Extension
• Domain-specific Acceleration
• RTL Integration
• Differential Testing

Innovation

• LLM-aided
• multi-agent
• RISC-V ISAX
• two-level IR
• compiler-agnostic formal verification