thewalrus 0.7.0

Open source library for hafnian calculation

The fastest exact hafnian library. For more information, please see the documentation.

Features

• The fastest calculation of the hafnians, loop hafnians, and torontonians of general and certain structured matrices.
• An easy to use interface to use the loop hafnian for quantum state calculations
• State of the art algorithms to sample from hafnian and torontonians of graphs.
• Efficient classical methods for approximating the hafnian of non-negative matrices.

Installation

Pre-built binary wheels are available for the following platforms:

	macOS 10.6+	manylinux x86_64	Windows 64bit
Python 3.5	✅	✅	✅
Python 3.6	✅	✅	✅
Python 3.7	✅	✅	✅

To install, simply run

pip install thewalrus

Compiling from source

The Walrus depends on the following Python packages:

• Python >=3.5
• NumPy >=1.13.3

In addition, to compile the C++ extension, the following dependencies are required:

• A C++11 compiler, such as g++ >= 4.8.1, clang >= 3.3, MSVC >= 14.0/2015
• Eigen3 - a C++ header library for linear algebra.

On Debian-based systems, these can be installed via apt and curl:

$ sudo apt install g++ libeigen3-dev

or using Homebrew on MacOS:

$ brew install gcc eigen

Alternatively, you can download the Eigen headers manually:

$ mkdir ~/.local/eigen3 && cd ~/.local/eigen3
$ wget http://bitbucket.org/eigen/eigen/get/3.3.7.tar.gz -O eigen3.tar.gz
$ tar xzf eigen3.tar.gz eigen-eigen-323c052e1731/Eigen --strip-components 1
$ export EIGEN_INCLUDE_DIR=$HOME/.local/eigen3

Note that we export the environment variable EIGEN_INCLUDE_DIR so that The Walrus can find the Eigen3 header files (if not provided, The Walrus will by default look in /use/include/eigen3 and /usr/local/include/eigen3).

Once all dependencies are installed, you can compile the latest stable version of the The Walrus library as follows:

$ python -m pip install thewalrus --no-binary :all:

Alternatively, you can compile the latest development version by cloning the git repository, and installing using pip in development mode.

$ git clone https://github.com/XanaduAI/thewalrus.git
$ cd thewalrus && python -m pip install -e .

OpenMP

libwalrus uses OpenMP to parallelize both the permanent and the hafnian calculation. At the moment, this is only supported on Linux using the GNU g++ compiler, due to insufficient support using Windows/MSCV and MacOS/Clang.

Using LAPACK, OpenBLAS, or MKL

If you would like to take advantage of the highly optimized matrix routines of LAPACK, OpenBLAS, or MKL, you can optionally compile the libwalrus such that Eigen uses these frameworks as backends. As a result, all calls in the libwalrus library to Eigen functions are silently substituted with calls to LAPACK/OpenBLAS/MKL.

For example, for LAPACK integration, make sure you have the lapacke C++ LAPACK bindings installed (sudo apt install liblapacke-dev in Ubuntu-based Linux distributions), and then compile with the environment variable USE_LAPACK=1:

$ USE_LAPACK=1 python -m pip install thewalrus --no-binary :all:

Alternatively, you may pass USE_OPENBLAS=1 to use the OpenBLAS library.

Software tests

To ensure that The Walrus library is working correctly after installation, the test suite can be run by navigating to the source code folder and running

$ make test

To run the low-level C++ test suite, Googletest will need to be installed. In Ubuntu-based distributions, this can be done as follows:

sudo apt-get install cmake libgtest-dev
cd /usr/src/googletest/googletest
sudo cmake
sudo make
sudo cp libgtest* /usr/lib/
sudo mkdir /usr/local/lib/googletest
sudo ln -s /usr/lib/libgtest.a /usr/local/lib/googletest/libgtest.a
sudo ln -s /usr/lib/libgtest_main.a /usr/local/lib/googletest/libgtest_main.a

Alternatively, the latest Googletest release can be installed from source:

sudo apt install cmake
wget -qO - https://github.com/google/googletest/archive/release-1.8.1.tar.gz | tar -xz
cmake -D CMAKE_INSTALL_PREFIX:PATH=$HOME/googletest -D CMAKE_BUILD_TYPE=Release googletest-release-1.8.1
make install

If installing Googletest from source, make sure that the included headers and libraries are available on your include/library paths.

Documentation

The Walrus documentation is available online on Read the Docs.

To build it locally, you need to have the following packages installed:

• Sphinx >=1.5
• sphinxcontrib-bibtex >=0.3.6
• nbsphinx
• Pandoc
• breathe >=4.12.0
• exhale
• Doxygen

They can be installed via a combination of pip and apt if on a Debian-based system:

$ sudo apt install pandoc doxygen
$ pip3 install sphinx sphinxcontrib-bibtex nbsphinx breathe exhale

To build the HTML documentation, go to the top-level directory and run the command

$ make doc

The documentation can then be found in the docs/_build/html/ directory.

Authors

Nicolás Quesada, Brajesh Gupt, and Josh Izaac.

All contributions are acknowledged in the acknowledgments page.

If you are doing research using The Walrus, please cite our paper:

Andreas Björklund, Brajesh Gupt, and Nicolás Quesada. A faster hafnian formula for complex matrices and its benchmarking on a supercomputer, Journal of Experimental Algorithmics (JEA) 24 (1), 11 (2019)

Support

• Source Code: https://github.com/XanaduAI/thewalrus
• Issue Tracker: https://github.com/XanaduAI/thewalrus/issues

If you are having issues, please let us know by posting the issue on our Github issue tracker.

License

The Walrus is free and open source, released under the Apache License, Version 2.0.