Generalised Circuit Partitioning for Distributed Quantum Computing

TL;DR

This paper presents a graph-based approach to optimize quantum circuit partitioning for distributed quantum computing, minimizing entanglement communication costs by jointly considering gate and state teleportation.

Contribution

It introduces a unified formulation for gate and state teleportation costs, enabling more efficient circuit partitioning in distributed quantum systems.

Findings

Improved e-bit cost over existing methods.

Enhanced scalability with genetic algorithm optimization.

Applicable to various circuit types.

Abstract

Distributed quantum computing (DQC) is a new paradigm aimed at scaling up quantum computing via the interconnection of smaller quantum processing units (QPUs). Shared entanglement allows teleportation of both states and gates between QPUs. This leads to an attractive horizontal scaling of quantum processing power, which comes at the expense of the additional time and noise introduced by entanglement sharing protocols. Consequently, methods for partitioning quantum circuits across multiple QPUs should aim to minimise the amount of entanglement-based communication required between distributed QPUs. Existing protocols tend to focus primarily on optimising entanglement costs for gate teleportation or state teleportation to cover operations between QPUs, rather than both at the same time. The most general form of the problem should treat gate and state teleportation on the same footing, allowing minimal cost circuit partitions through a combination of the two. This work introduces a graph-based formulation which allows joint optimisation of gate and state teleportation cost, including extensions of gate teleportation which group gates together for distribution using common resources. The formulation permits low e-bit cost for a variety of circuit types. Using a basic genetic algorithm, improved performance over state-of-the-art methods is obtained in terms of both average e-bit cost and time scaling.