ReQUSA: a novel ReRAM-based hardware accelerator architecture for high-speed quantum computer simulation
- 26 February 2024
- journal article
- research article
- Published by IOP Publishing in Physica Scripta
- Vol. 99 (3) , 035033
Abstract
Quantum computers are currently regarded as an emerging computing technology that can solve problems more quickly than classical computers. However, since constructing a general quantum computer is technically difficult, quantum computer simulation has been used instead of real quantum computers. Simulating quantum computers on classical computers is challenging because the time and resources required for the vectormatrix multiplication (VMM) increase exponentially with the number of qubits. This paper proposes a new accelerator architecture called ReQUSA that leverages resistive random access memory (ReRAM) to accelerate the quantum computer simulation. The ReQUSA employs a ReRAM crossbar array structure, which is specialized for implementing the VMM, and a realized state method for reduced VMM operation. To the best of our knowledge, ReRAM-based accelerators for quantum computer simulators has not been previously reported. Here, we describe the hardware design of the architecture and compare the performances (hardware resource, simulation time, and accuracy) of our accelerator with those of current quantum computer simulators (QuEST, QPlayer, and Qiskit). On average, our proposed architecture reduced the simulation times by factors of ×104 and ×103 (×102) on average from those of QuEST and QPlayer (also Qiskit), respectively. In addition, our architecture achieved 99% accuracy in 16-bit fixed-point data representation.Keywords
Funding Information
- Institute for Information & Communications Technology Planning & Evaluation (IITP) grant funding by the Korea government (No. 2020-0-00014)
- A Technology Development of Quantum OS for Fault-tolerant Logical Qubit Computing Environment) and the National Research Foundation of Korea (NRF) grant funded by the Korea government (RS-2023-00279770)
This publication has 35 references indexed in Scilit:
- GraphA: An efficient ReRAM-based architecture to accelerate large scale graph processingJournal of Systems Architecture, 2022
- CMOS-integrated memristive non-volatile computing-in-memory for AI edge processorsNature Electronics, 2019
- In-memory computing with resistive switching devicesNature Electronics, 2018
- Quantum machine learningNature, 2017
- Face classification using electronic synapsesNature Communications, 2017
- Improved Classical Simulation of Quantum Circuits Dominated by Clifford GatesPhysical Review Letters, 2016
- Massively parallel quantum computer simulatorComputer Physics Communications, 2007
- General-purpose parallel simulator for quantum computingPhysical Review A, 2002
- Quantum Mechanics Helps in Searching for a Needle in a HaystackPhysical Review Letters, 1997
- A special-purpose content addressable memory chip for real-time image processingIEEE Journal of Solid-State Circuits, 1992