Quantum Computing Achieves 70.72% Efficiency Gain with Novel Circuit Packing and Scheduling

The increasing demand for quantum computing resources creates bottlenecks as users face long waiting times for access to hardware, a problem stemming from current systems executing tasks one at a time. Miguel Palma, Shuwen Kan, and Wenqi Wei, from Fordham University, alongside Juntao Chen, Kaixun Hua from the University of South Florida, and Sara Mouradian from the University of Washington, address this challenge with a new approach to running multiple quantum circuits in parallel. Their work introduces CircPack, a framework specifically designed for trapped-ion quantum computers, which uniquely offer complete connectivity and high accuracy. By treating circuit scheduling as a packing problem that accounts for the physical constraints of these devices, CircPack demonstrably outperforms existing methods for other quantum platforms, achieving significantly higher fidelity, resource utilisation, and circuit simplification, and paving the way for scalable and efficient quantum cloud computing.

Quantum multi-programming (QMP) mitigates this issue by executing multiple circuits in parallel on a single device, but existing methods primarily target superconducting systems. These systems are characterised by limited connectivity, high crosstalk, and lower gate fidelity, presenting significant challenges for efficient multi-programming. This work focuses on trapped-ion architectures, which offer all-to-all connectivity, long coherence times, and high-fidelity mid-circuit measurement capabilities, providing a more suitable platform for advancing QMP techniques.

Circuit Packing and Scheduling for Trapped Ions

Scientists have developed CircPack, a novel framework that significantly enhances quantum multi-programming (QMP) on trapped-ion quantum computers. This work addresses limitations found in existing superconducting-based approaches and demonstrates substantial improvements in both hardware utilization and circuit throughput, paving the way for more efficient quantum cloud services.

The team achieved average qubit utilization values of 89.35% when processing a queue of 200 circuits, a considerable increase over previous methods. Furthermore, the framework reduced the number of circuit layers by 77.35%, streamlining computation and accelerating processing times.

Crucially, CircPack maintains high fidelity throughout this enhanced performance, achieving an average fidelity of 93.66% for multi-programmed circuits. Comparative evaluations reveal that CircPack outperforms other QMP methods by up to 70.72% in fidelity, 62.67% in utilization, and 32.80% in layer reduction factor.

CircPack Boosts Quantum Multiprogramming and Fidelity

This breakthrough is achieved by formulating circuit scheduling as a two-dimensional packing problem, specifically tailored to the unique characteristics of modular trapped-ion devices based on the Charge-Coupled Device (QCCD) architecture. The research also demonstrates the scalability of CircPack, successfully implementing balanced distributed scheduling across a simulated cluster of five quantum workers processing 3000 circuits. A minimal difference of only 2.83% was observed in makespan, and 1.74% for average utilization, between the most and least loaded workers, highlighting the framework’s ability to distribute workload effectively. These results confirm the potential of trapped-ion systems to improve the throughput of quantum cloud computing in the near future, offering a viable path towards scalable and efficient quantum computation.

Trap-Based Packing Boosts Quantum Throughput Significantly

CircPack addresses the growing need for efficient resource management in quantum cloud computing. By formulating circuit scheduling as a two-dimensional packing problem with hardware-specific constraints, CircPack achieves up to 70.72% better fidelity, 62.67% higher utilization, and 32.80% improved layer reduction compared to existing quantum multiprogramming approaches designed for superconducting systems.

The team’s results indicate that CircPack’s trap-based packing strategy and mechanisms for reducing ion transport contribute to these gains by minimizing accumulated noise. While achieving these improvements, the framework introduces a modest increase in processing time and CPU usage, a trade-off the researchers acknowledge. Future work will focus on evaluating CircPack on physical QCCD hardware and scaling the framework to handle larger circuits and longer queues. The researchers also plan to explore incorporating security and privacy constraints, generalizing the model to other all-to-all architectures like neutral atoms, and developing adaptive strategies for circuits exceeding trap size limitations. Additional research will investigate circuit transformations to further enhance compactness and noise resilience.