Hardening Tiny Tapeout Projects Locally

This document explains how to harden your Tiny Tapeout projects locally, to speed up iteration times. The whole process should take roughly 10 minutes.

It uses the factory-test project as an example.

Environment Setup

You need Python 3.11 or newer installed on your system. You can test which python version you have by running:

python3 --version

Note: On some systems, the python binary is called python and not python3.

We assume your project was cloned to ~/factory-test. If you don’t have a project yet, and want to follow these instructions to prepare your local setup, you can clone the factory-test repo by running the following command:

git clone https://github.com/TinyTapeout/tt08-factory-test ~/factory-test

1. Set up environment variables

export PDK_ROOT=~/ttsetup/pdk
export PDK=sky130A
export OPENLANE2_TAG=2.0.7

Note: the values of these values may change in the future - you can consult the tt-gds-action yaml for the latest values (look for the step named “Set up environment variables”)

2. Clone tt-support-tools

Clone the tt-support-tools repo (tt08 branch) inside the tt directory of your project:

cd ~/factory-test
git clone -b tt08 https://github.com/TinyTapeout/tt-support-tools tt

3. Python and Pip Dependencies

Make sure you have a recent version of python installed (3.11 or newer). Set up a virtual environment:

python3 -m venv ~/ttsetup/venv
source ~/ttsetup/venv/bin/activate

Then install the dependencies:

pip install -r ~/factory-test/tt/requirements.txt

Install the right version of openlane 2:

pip install openlane==$OPENLANE2_TAG

Note: On some systems, the python binary is called python and not python3.

4. Harden your project

Congratulations, you are ready to harden your project!

First, generate the openlane configuration file:

cd ~/factory-test
./tt/tt_tool.py --create-user-config --openlane2

Then run the following command to harden the project locally:

./tt/tt_tool.py --harden --openlane2

It’s also recommended to run the following command, checking for any synthesis / clock warnings:

./tt/tt_tool.py --print-warnings --openlane2


Once you set your environment, you can reharden at any time. Before running tt_tool.py, make sure to set the environment variables (as explained in step 1 above), and reactivate the Python virtual environment by running the following command:

source ~/ttsetup/venv/bin/activate

If you make changes to your project configuration (e.g. increase the number of tiles), you’ll need to update the openlane configuration file by running the following command in your project’s directory:

./tt/tt_tool.py --create-user-config --openlane2

To reharden, run:

./tt/tt_tool.py --harden --openlane2

Running the RTL tests

cd test
pip install -r requirements.txt
make -B

Running the gate level tests

cd test
pip install -r requirements.txt
TOP_MODULE=$(cd .. && ./tt/tt_tool.py --print-top-module)
cp ../runs/wokwi/final/pnl/$TOP_MODULE.pnl.v gate_level_netlist.v
make -B GATES=yes

Exporting the hardened design to a PNG file

Make sure you have the librsvg2-bin and pngquant packages installed:

sudo apt install librsvg2-bin pngquant

then you can use tt_tool to generate a render of the GDS as follows:

./tt/tt_tool.py --create-png --openlane2

The resulting optimised PNG file is called gds_render.png (but note that other gds_render* intermediate files are also left behind).

Speeding up Routing

For larger designs (or when close to maximum utilisation), routing can take a long time. If you have multiple CPU cores, you can try adding the following line to your config.json:


to speed things up (replace 8 with the number of CPU cores in your system). Note you should not commit this change to git.