CUNQA: a Distributed Quantum Computing emulator for HPC

TL;DR

CUNQA addresses the challenge of testing distributed quantum computing within HPC by emulating three DQC schemes using virtual QPUs. It introduces a three-layer stack (vQPUs, resource management, and a Python-based cunqa API) linked via a lightweight middleware, enabling on-node and co-located HPC integration. The paper demonstrates QPE across no-communication, classical-communication (IPEA), and quantum-communication models, showing φ converging to $1/\pi$ while revealing distinct performance profiles for each scheme. The work provides a practical, open-source testbed for exploring HPC-DQC software, scalability, and programming models ahead of real multi-QPU quantum hardware. This emulator lays groundwork for evaluating algorithms, scheduling, and inter-QPU protocols in realistic HPC environments.

Abstract

The challenge of scaling quantum computers to gain computational power is expected to lead to architectures with multiple connected quantum processing units (QPUs), commonly referred to as Distributed Quantum Computing (DQC). In parallel, there is a growing momentum toward treating quantum computers as accelerators, integrating them into the heterogeneous architectures of high-performance computing (HPC) environments. This work combines these two foreseeable futures in CUNQA, an open-source DQC emulator designed for HPC environments that allows testing, evaluating and studying DQC in HPC before it even becomes real. It implements the three DQC models of no-communication, classical-communication and quantum-communication; which will be examined in this work. Addressing programming considerations, explaining emulation and simulation details, and delving into the specifics of the implementation will be part of the effort. The well-known Quantum Phase Estimation (QPE) algorithm is used to demonstrate and analyze the emulation of the models. To the best of our knowledge, CUNQA is the first tool designed to emulate the three DQC schemes in an HPC environment.

Figures (12)

• Figure 1: All the possible integrations of QPUs in an HPC environment, a variation of the classification done by Elsharkawy et al. elsharkawy2023 that combines the two co-located models into one.
• Figure 2: Quantum software stack.
• Figure 3: All the possible DQC schemes.
• Figure 4: Quantum communication protocols extracted from the work of Barral et al. barral2024 with the communication qubits highlighted.
• Figure 5: In \ref{['subfig:virtual_qpu']} the vQPUs are responsible for the execution while in \ref{['subfig:qpus_executor']} the vQPUs delegate in an executor process.