Heterogeneous high-performance computing (HPC) systems offer novel architectures accommodating specialized processors that accelerate specific workloads. Near-term quantum computing technologies are poised to benefit applications as wide-ranging as quantum chemistry, machine learning, and optimization. A novel approach to scale these applications with future heterogeneous HPC is to enhance conventional computing systems with quantum processor accelerators. We present the eXtreme-scale ACCelerator programming model (XACC) to enable near-term quantum acceleration within existing conventional HPC applications and workflows. We design and demonstrate the XACC programming model within the C++ language by following a coprocessor machine model akin to the design of OpenCL or CUDA for GPUs. However, we take into account the subtleties and complexities inherent to the interplay between conventional and quantum processing hardware. The XACC framework provides a high-level API that enables applications to offload computational work represented as quantum kernels for execution on an attached quantum accelerator. Our approach is agnostic to the quantum programming language and the quantum processor hardware, which enables quantum programs to be ported to multiple processors for benchmarking, verification and validation, and performance studies. This includes a set of virtual numerical simulators as well as actual quantum processing units. The XACC programming model and its reference implementation may serve as a foundation for future HPC-ready applications, data structures, and libraries using conventional-quantum hybrid computing.
Recent advances in quantum computing have enabled the development of quantum processors with hundreds of qubits. However, noise continues to limit the amount of useful information that can be extracted from these systems, making it essential to identify the regime in which experimental outputs remain reliable. In this work, we benchmark Quantinuum Helios-1, a 98-qubit trapped-ion quantum processing unit, using the linear ramp quantum approximate optimization algorithm (LR-QAOA). To this end, we perform large-scale noiseless simulations on JUPITER, Europe's first exascale supercomputer, for circuits of up to 48 qubits and 3,384 two-qubit gates. These simulations, executed on 4,096 nodes equipped with 16,384 GH200 superchips and high-bandwidth CPU-GPU interconnects, provide a reference for validating experimental results at the edge of classical tractability. We find that, up to 48 qubits, Helios-1 remains in a noise-tolerant region, i.e., its samples cannot be clearly distinguished from those coming from a noiseless simulation. We then extend the analysis to larger system sizes using experimental data only, and apply a mean-of-means resampling procedure with a 3
σ threshold to determine whether the QPU output is statistically distinguishable from random sampling. This analysis identifies a regime of coherent performance up to 93 qubits (12,834 two-qubit gates), beyond which, at 95 qubits, the outputs become statistically indistinguishable from random sampling. These results demonstrate how exascale classical simulation can be used to validate quantum processors, and provide a quantitative boundary between noise-tolerant and random regimes in quantum processors.
Modular quantum processor architectures are envisioned as a promising solution for the scalability of quantum computing systems beyond the Noisy Intermediate Scale Quantum (NISQ) era. Based upon interconnecting tens to hundreds of quantum processors (i.e cores) via quantum coherent and classical links, this approach unravels the pressing limitations of densely qubit-packed monolithic processors, mainly by mitigating the requirements of qubit control and enhancing qubit isolation. Therefore, this new architectural design enables executing large-scale algorithms in a distributed manner. In order to assess the performance and optimize such architectures, it is crucial to analyze the quantum state transfers occurring via inter-core communication networks, referred to as inter-core qubit traffic. This would provide insights to improve the software and hardware stack for multi-core quantum computers. To this aim, we present a characterization of the spatio-temporal inter-core qubit traffic for different large-scale quantum algorithms. The programs are compiled on an all-to-all connected multi-core architecture that supports up to around 1000 qubits. We characterize the qubit traffic based on multiple performance metrics to assess the computational process and the communication overhead. Based on the showcased results, we conclude on the parallelization and scalability of presented algorithms, qualitatively evaluate the resource requirements as we scale circuit sizes, and lay the foundations of application-oriented benchmarking of large-scale multi-core architectures.
Noisy intermediate-scale quantum computers are widely used for quantum computing (QC) from quantum cloud providers. Among them, superconducting quantum computers, with their high scalability and mature processing technology based on traditional silicon-based chips, have become the preferred solution for most commercial companies and research institutions to develop QC. However, superconducting quantum computers suffer from fluctuation due to noisy environments. To maintain reliability for every execution, calibration of the quantum processor is significantly important. During the long procedure to calibrate physical quantum bits (qubits), quantum processors must be turned into offline mode. In this work, we propose a real-time calibration framework (RCF) to execute quantum program tasks and calibrate in-demand qubits simultaneously, without interrupting quantum processors. Across a widely used noisy intermediate-scale quantum (NISQ) evaluation benchmark suite such as QASMBench, RCF achieves up to 18% reliability improvement for applications. For reliability on different physical qubits, RCF achieves an average gain of 15.7% (up to 36.7%). For cloud quantum machines, the throughput can be improved up to 9.5 throughput per minute (6.5 on average) based on baseline calibration time. In conclusion, RCF offers a reliable solution for large-scale, long-serving quantum machines.