Q-Profile: Profiling Tool for Quantum Control Stacks applied to the Quantum Approximate Optimization Algorithm

Scaling up the number of qubits and accelerating quantum algorithms are important steps towards reaching quantum advantage. This poses heavy demands, as pulses need to be distributed to an increasing number of control channels and variational algorithms require rapid interleaving of quantum and classical computation. Assessing the bottlenecks in the control stack is therefore key to prepare it for reaching quantum advantage. However, existing benchmark suites suffer from indirect access to control hardware, limiting accuracy. In this work, we present Q-Profile, a tool to profile quantum control stacks using a direct connection from the host CPU to the control stack, providing fine accuracy in measuring runtime and allowing to identify performance bottlenecks. We demonstrate our tool by benchmarking the Quantum Approximate Optimization Algorithm on a Qblox Cluster for a virtual 4 to 14-qubit transmon system and identify the major execution bottlenecks. We estimate a 1.40x total runtime speedup by using an active qubit reset, and demonstrate a further 1.37x speedup by parallel control module initialization. The presented method of profiling is included in the open-source Quantify quantum control software, making it applicable to other control-stack providers and to other benchmarks, while still providing detailed information. By extension, this tool will enable identifying and eliminating bottlenecks for future quantum acceleration.