Comparison of Quantum Simulators for Variational Quantum Search: A Benchmark Study

TL;DR

This study benchmarks eight quantum simulators to identify the most efficient ones for running variational quantum search algorithms on many qubits, aiding future quantum algorithm development.

Contribution

It provides a comparative analysis of quantum simulators' performance for VQS, highlighting Pennylane with GPU and Qulacs as the most suitable options.

Findings

Pennylane with GPU and Qulacs are most efficient for VQS

Simulator performance degrades exponentially with qubit number

Results assist researchers in choosing appropriate simulators

Abstract

Simulating quantum circuits using classical computers can accelerate the development and validation of quantum algorithms. Our newly developed algorithm, variational quantum search (VQS), has shown an exponential advantage over Grover's algorithm in the range from 5 to 26 qubits, in terms of circuit depth, for searching unstructured databases. We need to further validate the VQS for more than 26 qubits. Numerous simulators have been developed. However, it is not clear which simulator is most suitable for executing VQS with many qubits. To solve this issue, we implement a typical quantum circuit used in VQS on eight mainstream simulators. Results show that the time and memory required by most simulators increase exponentially with the number of qubits and that Pennylane with GPU and Qulacs are the most suitable simulators for executing VQS efficiently. Our results aid researchers in selecting suitable quantum simulators without the need for exhaustive implementation, and we have made our codes available for community contributions.