Reproduce “VeriGraphi: A Multi-Agent Framework of Hierarchical RTL Generation for Large Hardware Designs” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Sazzadul Islam, Tasnim Tabassum, Hao Zheng
VeriGraphi is a multi-agent framework that generates synthesizable Verilog for large, hierarchical hardware designs directly from unstructured specifications by employing a spec-anchored Knowledge Graph. The framework demonstrated reduced human intervention and coding iterations compared to prior methods, while producing functionally correct and synthesizable RTL for complex designs such as the RISC-V 32I processor.
Key insights
- Explicitly capturing hardware hierarchy, interfaces, and inter-module relationships in a Knowledge Graph prior to code generation provides a machine-checkable blueprint that mitigates LLM hallucination and improves structural consistency.
- A progressive, multi-agent coding approach, guided by the detailed Knowledge Graph, enables efficient bottom-up RTL generation, enforcing interface consistency and dependency correctness at each submodule stage.
- Integrating a comprehensive, specification-driven verification module with PPA evaluation ensures end-to-end functional correctness and synthesizability, bridging high-level specifications to validated hardware.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.