Speaker
Description
Distributed quantum computing (DQC) is emerging as a practical route to scale quantum workloads beyond the qubit counts of individual devices, and its natural deployment target is the HPC centre, where quantum processing units are increasingly integrated alongside classical resources. However, the DQC software landscape remains fragmented: each network simulator, emulator and hardware stack exposes its own programming interface, so applications written for one platform cannot be reused on another, and results across platforms cannot be compared on equal terms.
We present NetQMPI, a communication library that brings the familiar MPI programming model to distributed quantum applications. NetQMPI provides point-to-point primitives built on entanglement distribution and quantum teleportation, together with collective operations implemented over shared GHZ states, allowing a single application to be expressed once in terms of ranks and communicators, independently of the underlying execution platform.
We describe the architecture of NetQMPI and its backend abstraction layer, and report on its integration with four distinct backends spanning different levels of abstraction: NetQASM/SquidASM, CUNQA, Qiskit Aer and Qoala. Using a common set of distributed benchmarks, we evaluate portability, fidelity and communication cost across these backends, showing that identical application code reproduces consistent results while exposing the trade-offs each platform makes between network realism and simulation performance.
Our results position NetQMPI as a portability layer for DQC in HPC environments, and as a practical tool for comparing DQC platforms under a unified programming interface.