Abstract
A closed-loop neuro-symbolic system automates the migration of Synopsys Design Constraints (SDC) from a reference project to a target project. A netlist abstraction layer parses netlists from the reference project and the target project by stripping combinational logic to extract registers, clocks, ports, and macros into semantic topological graphs. A structural mapper then matches equivalent hardware hierarchies across the topological graphs and detects topological differences, such as changes in register-to-register stage counts. A context-aware prompt builder combines the reference project SDC, hierarchical mappings, and topological differences into a structured prompt for a large language model (LLM) that generates a draft SDC. Within an agentic verification loop, an automated agent validates the draft SDC using a deterministic static timing analysis (STA) tool and parses any resulting STA error log to identify unmatched cells. To perform neuro-symbolic self-correction, the agent queries a netlist graph database to resolve synthesis-renamed registers and re-prompts the LLM with the resolved cell names until the STA tool produces a final verified SDC with zero errors, reducing timing-closure setup time and eliminating fabricated hardware paths.
Creative Commons License

This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
N/A, "System and Method for Automated Migration and Generation of Timing Constraints Using Graph-Aware Large Language Models", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/13165