Skip to content

C++17 Library for analysis, compilation/synthesis, and optimization of quantum circuits

License

Notifications You must be signed in to change notification settings

boschmitt/tweedledum

Repository files navigation

/!\ (Warning) If you have used tweedledum before: the master branch history is broken.
/!\ The new master branch is a completely rewrite of the library. The old version can be found
/!\ on **alpha** branch. (Sorry for the inconvenience!!---but it is for a great cause)

tweedledum is a library for synthesis, compilation, and optimization of quantum circuits. The library is written to be scalable up to problem sizes in which quantum circuits outperform classical ones. Also, it is meant to be used both independently and alongside established tools.

Its design is guided by three mantras:

  • Gotta run fast: execution-time performance is a priority.

  • Your compiler, your rules. You know better. At least, Tweedledum hopes so! The library provides a standard set of operators that can be easily extended (thanks to some type-erasure black magic). However, the library will leave your operators completely alone if you don't write passes that specifically manipulate them. Furthermore, Tweedledum will rarely take any decision in your behalf, i.e., it does not provide generic methods to optimize or synthesize circuits, you need to specifically call the algorithms you want.

  • Opinionated, but not stubborn. Many passes and synthesis algorithms have many configuration parameters. Tweedledum comes with reasonable defaults and curated opinions of what value such parameters should take. But in the end, it all up to you.

Corollary: Because of it's flexibility, Tweedledum is capable of accepting gates/operators that are defined as python classes. Indeed, any pythonic framework can use the library as a circuit manager. Meaning that the library can be used to slowly transition the core and performance sensitive parts of a pythonic framework to C++, while maintaining the capability of users to develop passes in python.

Installation

Known issues with macOS High Sierra (10.13) and macOS Mojave (10.14).

tweedledum has two python packages that can be installed using pip. For both, you will at least Python 3.6. The tweedledum package contains the latest stable release. You can install it from PyPI using:

  • Latest stable release (Linux/Mac/Windows)
pip install tweedledum

For the developers, users or researchers who are comfortable living on the absolute bleeding edge, tweedledum-dev contains that latest developments merged into the master branch.

  • Latest (Linux/Mac/Windows)
pip install tweedledum-dev

Warning: The two packages cannot be installed together.

Installation from source (Development)

Installing tweedledum from the source, instead of using the Python Package Index (PyPI) repository version, allows you to extend the latest version of the code. In the following, I will explain two workflows I personally use for development. Choose one that best suits your needs.

Alright, both workflows start the same way. You clone the repository:

git clone https://github.com/boschmitt/tweedledum.git

C++

The first workflow is pure C++. We start by creating a directory to hold the build output:

mkdir build      
cd build

Note that the library has a directory named examples/. If we set the TWEEDLEDUM_EXAMPLES CMake variable to TRUE. Any .cpp file in this directory will be compiled to its own executable.

So, lets assume you have a file named hello_world.cpp in the examples/. First we configure our project and enable the examples:

cmake -DTWEEDLEDUM_EXAMPLES=TRUE ..

If you are on a *nix system, you should now see a Makefile in the current directory. Now you can build the library by running make. At this point you can build the hello_world executable by calling

make hello_world

Once the examples have been built you can run it:

./examples/hello_world

C++ and Python

The second workflow is a bit of a hack. In Python we can install libraries in editable mode, meaning that code changes to the Python code in the project don't require a reinstall to be applied.

If you want to install it in editable mode, you can do this with:

pip install -e .

The only problem now, is that if we change the C++ code, we will need to reinstall the library. Fortunately, there is a way to circumvent this annoyance.

After installing in editable mode, you will see that in python/tweedledum/ there is a cpython shared library _tweedledum.cpython-... Remove this file:

rm python/tweedledum/_tweedledum.cpython-...

Now, we create a build directory as we did with the C++ workflow:

mkdir build      
cd build

We can manually build the cpython shared library using:

make _tweedledum

This will create the library in the build/ directory. Now, all we need to create a symbolic link in python/tweedledum/ that points the library in build/:

ln -s _tweedledum.cpython-39-darwin.so ../python/tweedledum/

Now, whenever we change the C++ code and rebuild the python library, the changes won't require a reinstall of the library to be available.

Used third-party tools

The library it is built, tested, bind to python, and whatnot using many third-party tools and services. Thanks a lot!

  • abc - ABC: System for Sequential Logic Synthesis and Formal Verification
  • bill - C++ header-only reasoning library
  • Catch2 test framework for unit-tests, TDD and BDD
  • CMake for build automation
  • Eigen template library for linear algebra
  • {fmt} - A modern formatting library
  • kitty - truth table library
  • lorina - C++ parsing library for simple formats used in logic synthesis and formal verification
  • mockturtle - C++ logic network library
  • nlohmann/json - JSON for Modern C++
  • parallel_hashmap - A family of header-only, very fast and memory-friendly hashmap and btree containers.
  • percy - C++ header-only exact synthesis library
  • pybind11 - Seamless operability between C++11 and Python
  • rang - A Minimal, Header only Modern c++ library for terminal goodies

Known issues

These are issues that hopefully will be fixed, but currently are unsolved. If you know how to help with one of these issues, contributions are welcome!

macOS: High Sierra (10.13) and Mojave (10.14)

tweedledum offers limited support for both systems. While wheels might be available for some releases, it is strongly advised to install tweedledum or tweedledum-dev by building them directly from source. For example:

CC=gcc-10 CXX=g++-10 CXXFLAGS="-static-libgcc -static-libstdc++" pip install tweedledum --no-binary :all:

Note that such command requires a working gcc10 installation. (It also works with gcc11, but no further tests were made.) I recommend the use of Homebrew to install gcc. (Or maybe Tigerbrew)

License

This software is licensed under the MIT licence (see LICENSE).

EPFL logic synthesis libraries

tweedledum is part of the EPFL logic synthesis libraries. The other libraries and several examples on how to use and integrate the libraries can be found in the logic synthesis tool showcase.