Speaker
Description
CUNQA: a Distributed Quantum Computing emulator for HPC
Two trends are shaping the future of quantum computing. On the one hand, hardware roadmaps point to distributed quantum computing (DQC), in which several quantum processing units (QPUs) are linked to act as a larger machine. Meanwhile, on the other hand, HPC centers are moving towards integrating QPUs as accelerators alongside GPUs and FPGAs. The infrastructures that combine both, clusters of interconnected QPUs embedded in HPC systems, do not exist yet, but the algorithms, programming models and workflows that will run on them should be explored now. Existing tools cover only half of the picture: HPC-quantum integration frameworks treat QPUs as isolated devices with no communication between them, while quantum-network simulators model communication in detail but ignore the HPC environment.
CUNQA is an open-source emulator designed to close this gap. Its objective is to let researchers and developers deploy, program and study DQC infrastructures inside a real HPC environment before the corresponding hardware is available. To do so, it supports the three DQC communication models: no communication, where independent QPUs run separate tasks in parallel; classical communication, where QPUs exchange classical bits at runtime; and quantum communication, where QPUs share entanglement to perform non-local operations.
The tool is built around a clear division: the infrastructure is emulated, while the physics is simulated. QPUs are represented as virtual QPUs (vQPUs): HPC processes with their own classical resources, which users launch, query and address exactly as they would with a real QPU. The management of the quantum state is delegated to interchangeable simulators, so CUNQA focuses on reproducing how a DQC system is structured and used in an HPC environment rather than on simulation itself. Regarding the HPC context, CUNQA is consistent with it, leaving the resource management in the user's hands and using SLURM-based commands to deploy the vQPUs, letting users configure the simulated devices, the simulator, the communication model and how the QPUs integrate with the HPC system. On top of this, a Python interface adds the directives that distributed algorithms need: classical send and receive, entanglement generation, and ready-made quantum communication protocols.
To the authors' knowledge, CUNQA is the only tool combining HPC integration with all three DQC communication models. It is intended not as a finished product but as an exploratory platform for understanding how DQC will fit into the HPC systems of the coming years.