Multithreaded parallelism for heterogeneous clusters of QPUs

TL;DR

MILQ tackles the challenge of executing batches of quantum circuits on heterogeneous backends by combining circuit knitting to fit circuits to device capacities with a MILP-based scheduler that minimizes the makespan $c_{\max}$. It introduces a modular architecture (QPU wrapper, scheduler, compilation pipeline) and demonstrates performance gains against a baseline, achieving up to ~26% improvement in some configurations. The work highlights practical considerations such as hardware timing uncertainty, planning for online rescheduling, and the need for scalable heuristics, positioning MILQ as a versatile runtime component for multi-backend quantum computing. Overall, MILQ offers an end-to-end approach to improve hardware utilization in near-term quantum infrastructures and provides a foundation for integrating circuit-knitting decisions directly into scheduling.

Abstract

In this work, we present MILQ, a quantum unrelated parallel machines scheduler and cutter. The setting of unrelated parallel machines considers independent hardware backends, each distinguished by differing setup and processing times. MILQ optimizes the total execution time of a batch of circuits scheduled on multiple quantum devices. It leverages state-of-the-art circuit-cutting techniques to fit circuits onto the devices and schedules them based on a mixed-integer linear program. Our results show a total improvement of up to 26 % compared to a baseline approach.

Figures (4)

• Figure 1: Circuit knitting in two variants: space-like and time-like cuts are possible.
• Figure 2: Simplified overview of the intended workflow.
• Figure 3: Schedule of the sample problem obtained by the simple schedule.
• Figure 4: Makespan results for the three configurations baseline, simple and extended in two different settings.