This repository proposes a set of five configurable quantum circuits for evaluation of Nvidia's CUDA-Q simulator backends, targeting specifically plain Tensor Network and Matrix Product State (MPS) simulators.
To obtain help on the utilization of the benchmark script, use:
python benchmark.py --help
Depending on the architecture of the executing system - and notably on Arm platform -,
it might be simpler to execute CUDA-Q within a Docker container. Nvidia provides a
Docker image
which streamlines this process. We propose an experimental script scripts/run.sh, which
automatizes creating and running the container, used as a wrapper:
scripts/./run.sh python3 benchmark.py --help
G. Schieffer, S. Markidis, and I. Peng. (2025). Harnessing CUDA-Q's MPS for Tensor Network Simulations of Large-Scale Quantum Circuits. 2025 33rd Euromicro International Conference on Parallel, Distributed and Network-Based Processing.
Arxiv pre-print: https://doi.org/10.48550/arXiv.2501.15939