Skip to content

ORNL-QCI/tnqvm

Repository files navigation

Branch Status
master Linux CI

TNQVM Tensor Network XACC Accelerator

These plugins for XACC provide an Accelerator implementation that leverages tensor network theory to simulate quantum circuits.

Installation

With the XACC framework installed, run the following

$ mkdir build && cd build
$ cmake .. -DXACC_DIR=$HOME/.xacc (or wherever you installed XACC)
$ make install

TNQVM can be built with ExaTN support, providing a tensor network processing backend that scales on Summit-like architectures. To enable this support, first follow the ExaTN README to build and install ExaTN. Now configure TNQVM with CMake and build/install

$ mkdir build && cd build
$ cmake .. -DXACC_DIR=$HOME/.xacc -DEXATN_DIR=$HOME/.exatn
$ make install

To switch tensor processing backends use

auto qpu = xacc::getAccelerator("tnqvm", {std::make_pair("tnqvm-visitor", "exatn")});

or in Python

qpu = xacc.getAccelerator('tnqvm', {'tnqvm-visitor':'exatn'})

MPI Execution

TNQVM's exatn-mps visitor can support multi-node execution via MPI.

Prerequisites: ExaTN is built with MPI enabled, i.e., setting MPI_LIB and MPI_ROOT_DIR when configuring the ExaTN build.

To enable MPI in TNQVM, add -DTNQVM_MPI_ENABLED=TRUE to CMake along with other configuration variables.

A simulation executable which uses the exatn-mps visitor, e.g. via

auto qpu = xacc::getAccelerator("tnqvm", { std::make_pair("tnqvm-visitor", "exatn-mps")});

can be executed with MPI using mpiexec -np <number of processes> <executable>.

Documentation

Questions, Bug Reporting, and Issue Tracking

Questions, bug reporting and issue tracking are provided by GitHub. Please report all bugs by creating a new issue with the bug tag. You can ask questions by creating a new issue with the question tag.

License

TNQVM is licensed - BSD 3-Clause.

Cite TNQVM

If you use TNQVM in your research, please use the following citation

@article{tnqvm,
    author = {McCaskey, Alexander AND Dumitrescu, Eugene AND Chen, Mengsu AND Lyakh, Dmitry AND Humble, Travis},
    journal = {PLOS ONE},
    publisher = {Public Library of Science},
    title = {Validating quantum-classical programming models with tensor network simulations},
    year = {2018},
    month = {12},
    volume = {13},
    url = {https://doi.org/10.1371/journal.pone.0206704},
    pages = {1-19},
    number = {12},
    doi = {10.1371/journal.pone.0206704}
}