Workshop: The 2nd International Workshop for Software Frameworks and Workload Management on Quantum-HPC Ecosystems
Authors: Sangram Deshpande (North Carolina State University, Oak Ridge national lab); Sharmin Afrose and Nicholas Stapleton (Oak Ridge National Laboratory); Nirajan Banjade (Texas state University); Romit Ghosh (Rutgers University); and Vicente Leyton-Ortega, Travis Humble, and Tirthankar Ghosal (Oak Ridge National Laboratory)
Abstract: This paper presents an end-to-end workflow combining a locally hosted LLM for circuit generation, a deterministic stabilizer-code module for QEC encoding, and Constraint-Aware Hardware Graph Retrieval for identifying suitable connected hardware regions. We evaluate six compilation strategies across 720 method–configuration combinations, covering six IBM calibration snapshots, five three-qubit circuit families, and four optimization objectives. Compared with unoptimized raw mapping, the multi-candidate hardware_aware strategy reduces mean circuit depth from 956 to 468 and two-qubit gate count from 657 to 268. Compared with full-device Sabre, improvements are smaller, reducing depth from 488 to 468 and two-qubit gates from 281 to 268. Greedy CAHGR alone underperforms conventional baselines, demonstrating that calibration-aware region selection must be combined with effective placement and routing. Conclusions are limited to compilation efficiency because execution quality uses a no-fault proxy rather than logical fidelity or hardware execution.
Back to The 2nd International Workshop for Software Frameworks and Workload Management on Quantum-HPC Ecosystems Archive Listing
Back to Full Workshop Archive Listing