CircuitGraph

CircuitGraph is a library for working with hardware designs as graphs. CircuitGraph provides an interface to do this built on NetworkX, along with integrations with other useful tools such as sat solvers and the Yosys synthesis tool, and input/output to verilog.

Overview

The Circuit class is at the core of the library and it is essentially a wrapper around a NetworkX graph object. This graph is accessable through the graph member variable of Circuit and can be used as an entrypoint to the robust NetworkX API.

Here's a simple example of reading in a verilog file, adding a node to the graph, and writing back to a new file.

import circuitgraph as cg

c = cg.from_file('/path/to/circuit.v')
# Add an AND gate to the circuit that takes as input nets o0, o1, o2, o3
c.add('g', 'and', fanin=[f'o{i}' for i in range(4)])
cg.to_file(c, '/path/to/output/circuit.v')

The documentation can be found here.

Installation

CircuitGraph requires Python3.7 or greater The easiest way to install is via PyPi:

pip install circuitgraph

To install from the release, download and:

pip install circuitgraph-<release>.tar.gz

Finally, to install in-place with the source, use:

cd <install location>
git clone https://github.com/circuitgraph/circuitgraph.git
cd circuitgraph
pip install -r requirements.txt
pip install -e .

Optional Packages

In addition to the packages enumerated in requirements.txt, there are a few tools you can install to enable additional functionality.

If you would like to use the satisfiability functionality, install PySAT.

Open source synthesis can be perofmred by installing Yosys and adding it to your path. Alternatively, Genus or DesignCompiler can be used by providing the path to a generic library to use by setting the CIRCUITGRAPH_GENUS_LIBRARY_PATH and CIRCUITGRAPH_DC_LIBRARY_PATH environment variables.

Contributing

If you have ideas on how to improve this library we'd love to hear your suggestions. Please open an issue. If you want to develop the improvement yourself, please consider the information below.

Coverage is computed using Codecov. If you would like to generate coverage information locally, install coverage and codecov.

pip install coverage codecov
make coverage

Documentation is built using pdoc3.

pip install pdoc3
make doc

Tests are run using the builtin unittest framework. Some basic linting is performed using flake8.

pip instsall flake8
make test

Code should be formatted using black. Pre-commit is used to automatically run black on commit.

pip install black pre-commit
pre-commit install

Pre-commit also runs a few other hooks, including a docstring formatter and linter. Docs follow the numpy documentation convention.

Citation

If you use this software for your research, we ask you cite this publication: https://joss.theoj.org/papers/10.21105/joss.02646

@article{sweeney2020circuitgraph,
  title={CircuitGraph: A Python package for Boolean circuits},
  author={Sweeney, Joseph and Purdy, Ruben and Blanton, Ronald D and Pileggi, Lawrence},
  journal={Journal of Open Source Software},
  volume={5},
  number={56},
  pages={2646},
  year={2020}
}

Acknowledgements

Circuitgraph icon designed by ncasti.