Abstract
The technology described in this paper relates to transformation of functional register-transfer level (RTL) hardware designs for implementation by electronic design automation (EDA) physical synthesis tools. An intelligent, autonomous, and self-correcting RTL transformation system that is based on large language models (LLMs) is proposed. The system includes an LLM agent that is based on one or more LLM models, and a robust closed-loop transformation validation agent that is integrated with the LLM agent. A continuous feedback loop is used to allow the agent for RTL transformation validation to provide feedback to the LLM agent as error messages. The LLM interprets the error messages received from the agent for RTL transformation validation. The LLM agent may also interpret error messages received from other validation tools, e.g., validation tool linting. The LLM agent uses the feedback from various sources to automatically and iteratively revise the generated RTL until no errors are reported.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
N/A, "Large Language Model (Llm) – Agent System For Register Transfer Level (Rtl) Transformation", Technical Disclosure Commons, (February 03, 2026)
https://www.tdcommons.org/dpubs_series/9270