NVIDIA Quantum Benchmarks v0.6.0

[fmozafari]

We are excited to announce the release of NVIDIA Quantum Performance Benchmark Suite v0.6.0 🚀 , bringing new features and improvements for quantum simulation benchmarking.

New Changes

• Set default cudaq simulator to CPU (qpp-cpu).
• Explicitly set default fuse sizes for cudaq to None, therefore cuquantum chooses the best fuse size based on the architecture.
• Update the other backends than cudaq to auto select fuse size to the best of their ability. The only backend that can not do that right now is Qsim and the related family.
• Installing the benchmarking repository inside the CUDA Quantum container is not as simple as pip install ., documenting it in the README as a point of reference.
• Fix the code to also run on CPU-only containers.

Compatibility note

• Support Pennylane version 0.42.0. Fix ControlledQubitUnitary syntax for the new version. Remove the support of very old versions <0.33.0. Support MPI to run with multi-GPU for lightning-kokkos and lightning-gpu

Installation Instructions

Same as in the previous release, it can be easily installed by cloning this repository as follows:

$ git clone https://github.com/NVIDIA/cuQuantum.git
$ cd cuQuantum/benchmarks/
$ pip install .
$ nv-quantum-benchmarks --help

This benchmark suite works either without or with our cuQuantum Appliance container. For further information, please see the benchmark project README.

Happy benchmarking with NVIDIA Quantum Performance Benchmark Suite v0.6.0!