Add Wakeel submission — ISSCC27 Code-a-Chip

[Rishikesavan0129]

Wakeel — Self-Healing, Frequency-Aware Configuration Engine for Open-Source RTL-to-GDSII Flows

Author: Rishi Kesavan
Affiliation: Mahendra College of Engineering, Salem, Tamil Nadu, India (B.E. ECE)
Category: Educational flow built with open-source tools and PDKs / exploration of open-source PDK-driven design-space behavior
Path: ISSCC27/submitted_notebooks/wakeel/
License: Apache License 2.0

What this is

Wakeel is an orchestration layer on top of OpenLane (SkyWater sky130) and OpenROAD-flow-scripts (ASAP7 7nm) that encodes tacit RTL-to-GDSII expertise as executable, inspectable logic instead of leaving it in an engineer's head. Three pieces, all demonstrated in the notebook against real code from the live project (not a simplified reimplementation):

1. Schema-driven configuration generation — every OpenLane/ORFS parameter lives in one JSON schema; adding a new parameter is a one-line JSON edit, never a code change.
2. Frequency-aware physical profiling — synthesis/resizer aggressiveness escalates automatically with target clock frequency and PDK, instead of one static default that's wrong at either extreme.
3. Self-healing knowledge base — reads real flow-failure signoff content (not just an exit code) and proposes a diagnosed, schema-grounded configuration fix.

Real hardware validation

Part 4 includes a real ORFS/ASAP7 synth_power_only sweep (wakeel_alu, 0.2–2.0 GHz, 5 points, all SUCCESS) run locally with Docker + the ASAP7 PDK — not synthetic numbers. The CSV backing that result is included in ppa_reports/ alongside the notebook.

Reproducibility

The notebook clones the project at a pinned commit (not main), so Parts 1–3 run standalone in Colab with only matplotlib/numpy/pandas. Part 4's real hardware run needs a local Docker + PDK install — the notebook explains exactly why and how to regenerate it, per the program FAQ's guidance for this case.

Full project: https://github.com/Rishikesavan0129/wakeel