Executing Applications in an Hybrid HPC-Quantum Environment: Integration and Benchmark

Andres Gomez Tato, head of applications and projects at the Galician Supercomputing Center (CESGA), describes how CESGA integrated a quantum computer with its HPC systems and what benchmarking such hybrid environments requires. He frames current quantum devices as immature analog prototypes spanning many competing technologies, notes that all quantum algorithms are hybrid with classical pre- and post-processing, and positions QPUs as accelerators executing small kernels within larger workflows. He outlines three deployment models, from standalone cloud-accessed systems through loosely Ethernet-coupled integration to future tightly integrated QPUs, and three user levels from high-level programmers to pulse-level experimentation. CESGA's user base spans proof-of-concept research (including a molecular optimization project on beer freshness with Fujitsu run on real quantum hardware), device testing, tool integrators, and large training events serving over one hundred simultaneous students. He argues the community needs reproducible, agnostic, verifiable benchmarks covering the full hardware and software stack, with classically checkable solutions. On CESGA's QMIO infrastructure, a QPU attached to a single HPC node, latency measurements show more than three orders of magnitude between circuit execution time and input processing time, with parser performance varying widely across OpenQASM and QIR representations, indicating substantial headroom for software optimization. Iterative and variational algorithms that require thousands to millions of circuit executions make throughput and latency central requirements. Future work includes an emulator of distributed quantum infrastructures with virtual QPUs, collaborations with companies developing parallel quantum execution tools, and a commitment to releasing everything open source.