Core Concepts
Quantum program scheduling can significantly reduce the execution latency of quantum programs on superconducting quantum processors by considering circuit width, shot number, and submission time.
Abstract
The content discusses the Quantum Program Scheduling Problem (QPSP) to improve the efficiency of executing quantum programs on superconducting quantum processors.
Key highlights:
- Quantum programs are often executed serially on current quantum processors, leading to long queue times and low qubit utilization.
- The authors formulate the QPSP to minimize the execution latency while maintaining high fidelity and fairness.
- A novel scheduling method is proposed that considers the circuit width, shot number, and submission time of quantum programs to determine the execution order.
- Three greedy baseline methods are also devised for comparison.
- Extensive experiments on a simulated noise model and a real superconducting quantum processor (Xiaohong) show that the proposed method significantly reduces the QPU time and turnaround time compared to the default serial execution and the greedy baselines, with a small cost in fidelity.
- The runtime overhead of the scheduling algorithm is small, indicating its scalability on large quantum processors.
Stats
The average number of pending jobs on IBM Perth is about 2,540, and the average queue time is about 6.7 hours.
The range of circuit width for the noise model and Xiaohong is [3, 16] and [6, 6.65] on average, respectively.
The range of circuit depth is [5, 99] and [21, 98] on average, respectively.
Quotes
"Though there exist some quantum cloud services, the growing need for quantum hardware outpaces the open access to quantum hardware."
"Multi-programming on quantum processors is a complicated task. The execution order of programs will affect the performance of multi-programming."