Presentation Information
[C-12-38]AI-Driven Portable Digital Design Flow: Text-to-GDS
◎△WENQIAN WANG1, Haruto Tanaka1, Tianshi Hu1, Kanta Fukuda1, Nabil Soliman1, Ned Sureechainirun1, Yuncheng Zhang1, Hiroyuki Sakai1, Kazuaki Kunihiro1, Kenichi Okada1 (1. Institute of Science Tokyo)
Keywords:
digital circuit,AI agent,electronic design automation (EDA),Large language model (LLM)
Large language models (LLMs) can write hardware code well, but cannot yet drive the long RTL‑to‑GDSII backend reliably. This work presents an AI‑driven, portable text‑to‑GDS flow, where a router orchestrates decoupled skills wrapping commercial EDA tools, so that LLMs and tools can be swapped easily, and an agentic diagnosis loop fixes each error at a suitable entry point. As examples, a 16×32 asynchronous FIFO and a 32‑bit floating‑point adder are produced from simple text.
