EDATracer: Agentic Framework for EDA Artifact Analysis

A novel framework named EDATracer has been developed by researchers to facilitate the large-scale examination of electronic design automation (EDA) artifacts. This framework tackles the complexities involved in analyzing diverse artifacts, including source files, scripts, logs, netlists, and reports, which are essential for debugging, optimization, and comprehending design processes. By structuring these artifacts into a knowledge graph alongside a semantic vector index, EDATracer allows LLM agents to effectively gather evidence from various artifact types. To bolster the framework, the researchers assembled an 18.9 GB dataset featuring 2,787 synthesizable open-source chip designs. This initiative seeks to enhance evidence-based reasoning in EDA support, addressing the shortcomings of current LLM-based methods that lack public benchmarks for cross-artifact evaluation.