Accelerate Silicon Design with Jupyter Notebooks
summary
A reproducible flow turns a high‑level DSL (XLS/dslx) into GDS using SkyWater Sky130 PDK and GF 180 open PDKs, Yosys/XLS for RTL, OpenROAD/OpenLane for P&R, Magic/CALIBER for GDS sign‑off, all packaged as Jupyter+Conda notebooks that run on Colab or cloud. A terraform repo rad-lab provisions notebook clusters; optimization uses Google Vizier (open source) to run thousands of parameterized runs (examples show ~10k) that quickly converge. Each job emits a standalone notebook; Open MPW batches accept ~40 open designs and Google covers mask and wafer costs for accepted open‑source submissions.
AI-written summary · about
chapters (24)
- 0:00Intro
- 0:31Who are you
- 1:07Mission at Google
- 1:32Vision
- 1:59Open Design Elements
- 7:11Software Chain
- 9:01Packages
- 9:53Sharing Notebooks
- 10:47Demo
- 13:26Hardware Synthesis
- 16:07Hardware Synthesis Example
- 19:14Open Lane
- 19:45Configuration
- 20:44PIN
- 23:17Floor Plan
- 25:50Routing
- 27:33Signoff
- 29:01Multiplication
- 32:03Open Link
- 32:40Configuration Options
- 34:18Preview
- 36:54Visualization
- 38:39Aggregate
- 39:05OpenMPW Shutter
related talks
01h 23mImproving Your Data Modeling Work Through Open-Source Software
42mSimplifying Data Analysis with GitHub Codespaces, Jupyter Notebooks & Open AI
11mDemo: Accelerating SW Development and Validation of RISC-V Based Designs
37mReproducible and shareable notebooks across a data science team
31mEvaluating fire simulators using Docker, Dask and Jupyter notebooks!
36mThe Lifecycle of a Jupyter Environment
49mExperimenting with Large Data Volumes Using Notebooks
18mBeyond EDA lies Edalize