EdgeEDA Agent

EdgeEDA is an agentic design-space tuner for OpenROAD-flow-scripts (ORFS). It runs multi-fidelity experiments (synth/place/route/finish), computes a scalar reward, and iteratively proposes knob settings.

What is included

  • Agent code under src/edgeeda
  • CLI for running sweeps and exporting plots
  • Example configs under configs/
  • Utility scripts under scripts/

What you need locally

  • OpenROAD-flow-scripts (ORFS)
  • OpenROAD + Yosys available on PATH, or set OPENROAD_EXE / YOSYS_EXE

Quickstart

python -m pip install -r requirements.txt

# Run a quick sweep (edit config paths as needed)
python -m edgeeda.cli tune --config configs/gcd_nangate45_sweep24_finish.yaml --timeout 900

# Generate plots
python generate_plots.py --db runs/sweep_finish.sqlite --out runs/sweep_finish/plots

Notes

This repo does not ship ORFS artifacts or large result folders to keep size manageable. Use the provided configs to point to your ORFS checkout.

Citation

If you use this agent in academic work, cite the accompanying paper in IEEE_EdgeEDA_Agent_ISVLSI/ of the source repository.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support