OpenROAD and the Era of Agents
summary
OpenROAD becomes an experimental substrate where agents edit code, tune flows, and run closed-loop R&D: OpenROAD hosts repo-scale agents like Autopeda that edit, compile, and benchmark; flow agents such as ORFS agent reach similar QOR with ~40% fewer iterations and beat Bayesian tuning by ~13% on wire length/timing; code-evolving agents like Helix and router-focused GR Evolve change placer/router heuristics; partition tweaks (TritonPart → FastPart) find world‑best cuts and give up to 10–100× speedups.
AI-written summary · about
chapters (10)
- 0:00OpenROAD and agents intro
- 1:07Agents transforming EDA
- 2:54Agents in OpenROAD types
- 4:38Agentic EDA research frontiers
- 6:27Agentic improvements case studies
- 12:03Reliable agentic EDA challenges
- 13:48Human roles and IP in agents
- 14:51Governance and ecosystem challenges
- 16:14OpenROAD as agentic substrate
- 17:45Q&A on agentic EDA
related talks
01h 00mOpenROAD: The journey so far and roadmap
26mAgent Omelets & Other Forbidden Techniques
17mSoftware Development Agents: What Works and What Doesn't
33mAutomatic Agent Optimization and the Rise of Self-Improving Agents
25mBeyond Orchestration: An Open Source Architecture for Agentic Systems
25mBeyond the IDE: Toward Multi-Agent Orchestration
25mDirecting a Swarm of Agents for Fun and Profit
16mBeyond the IDE: The Future of Developing Software With Agents