AMD and OpenQSE Advance Quantum-HPC Integration at SC26
Home
Latest reporting
News
AMD and OpenQSE unveil a modular software stack for hybrid quantum-HPC workflows at Supercomputing 2026, emphasizing vendor-neutral standards.
AMD, in collaboration with the Open Quantum-HPC Software Ecosystem (OpenQSE), has introduced a modular software stack aimed at integrating quantum computing with high-performance computing (HPC). The announcement was made at Supercomputing 2026 (SC26), where ecosystem members demonstrated a reference implementation that connects independently developed quantum-HPC software and hardware components into a functional, testable stack.
Hybrid Workflow Models for Quantum-HPC Integration
The OpenQSE reference implementation supports two hybrid workflow models: staged workflow orchestration and co-scheduled hybrid execution. These models are designed to test integration boundaries across workflows, scheduling, compilation, resource interfaces, and device-facing runtimes.
In the staged workflow model, classical and quantum resources are reserved and released per stage. Classical preprocessing, quantum execution, and postprocessing occur sequentially, with data movement between stages managed by workflow engines such as Tierkreis, StreamFlow, and Prefect. This approach is suited for workflows where classical and quantum tasks can be decoupled in time.
In contrast, the co-scheduled hybrid execution model reserves classical and quantum resources for the same execution window. A hybrid request is managed by classical workload managers like Slurm, with quantum-side capacity evaluated through adapters such as qfw-slurm. This model enables synchronized execution of classical and quantum tasks, emphasizing tighter integration.
Technical Architecture and Vendor Neutrality
The OpenQSE stack is structured as a layered execution path, connecting application workflows to quantum resources through explicitly testable interfaces. Core components include:
Workflow orchestration systems (e.g., Tierkreis, StreamFlow, Prefect) for sequencing stages and managing data movement.Classical resource managers like Slurm, with Flux targeted for future integration.Quantum resource interfaces, such as QFw, qhw-admission, and qhw-scheduler, which handle quantum task reservation, capacity evaluation, and dispatch.Compiler tools like MQT Core, Qiskit, and PennyLane, which ensure portability across quantum hardware while maintaining vendor neutrality.
The stack also incorporates a common Quantum Resource Interface shim, bridging orchestration layers with heterogeneous quantum resources. Current implementations support quantum hardware from IBM, IQM, Pasqal, and Amazon Braket, with Munich Quantum Valley contributing to federated setups across LRZ and AWS.
Future Directions and Collaborative Development
Looking ahead, the OpenQSE initiative is exploring architectures for fault-tolerant quantum computing (FTQC) and quantum error correction (QEC). Efforts involving Backline, co-developed by Xanadu and AMD, and SpaceRT focus on real-time quantum-classical integration, addressing low-latency execution constraints critical for future FTQC models.
The OpenQSE project is a collaborative effort involving national laboratories, industry leaders, cloud providers, quantum hardware vendors, and academic institutions. Key contributors include Oak Ridge National Laboratory (ORNL), IBM, Quantinuum, AMD, AWS, and Munich Quantum Valley. By fostering multi-organizational participation, OpenQSE aims to establish vendor-neutral standards for hybrid quantum-HPC systems, ensuring broad applicability and interoperability.
