Workshops Archive

Dispatch Granularity in Quantum-HPC ADMM Optimization: Implications for Near-Real-Time Computing


Workshop: The 2nd International Workshop for Software Frameworks and Workload Management on Quantum-HPC Ecosystems

Authors: Giuliana Siddi Moreau, Manuela Profir, Maria Laura Clemente, Lorenzo Pisani, and Lidia Leoni (CRS4)

Abstract: Iterative-consensus decomposition enables combinatorial optimization problems exceeding a single quantum processing unit (QPU) by solving coupled block subproblems and reconciling them at a consensus barrier. We evaluate whether sub-solver quality or execution mode governs performance. Across 880 byte-identical sub-QUBOs of widths 6–13, non-variational digitized counterdiabatic quantum optimization (DCQO) and variational QAOA match exact enumeration at widths ≤ 7 under equal per-circuit shot budgets, whereas classical simulated annealing produces materially larger gaps. When routed onto the measured coupling map of the 54-qubit IQM Emerald, DCQO and QAOA circuit depths differ by at most 9% at supported widths. Their principal distinction is dispatch granularity: DCQO requires one circuit per block and permits round-level batching, whereas QAOA requires 25 sequential circuit evaluations per block in our execution model. Thus, for near-real-time hybrid optimization, execution-mode selection matters more than sub-solver choice.


Back to The 2nd International Workshop for Software Frameworks and Workload Management on Quantum-HPC Ecosystems Archive Listing

Back to Full Workshop Archive Listing