Institute for Advanced Simulation (IAS)

Jülich Supercomputing Centre (JSC)

Integrating quantum computers into HPC environments

Within the next decade, quantum computers (QC) will likely not be able to solve real world problems faster than conventional high-performance computers (HPC). However, hybrid classical-quantum systems, e.g. by using QC as accelerators, such as GPUs are currently used, show great potential to provide a speed-up already during the next 10 years. This does not imply that hybrid systems will be outdated once QC deliver benefits as stand-alone system, since it is expected that there will still be problems to be solved where work partitioning between classical and QC will be most efficient.

Hybrid calculations require a stable connection between the QC and the HPC system and the workflow between the systems must be organized by tailored software. The current challenge is that no linux-like software exists for QC, which serves as a unified interface on top of different software and firmware stacks of the different quantum devices. This implies the need to develop a full software stack for every single QC system that handles the workflow in the hybrid environment.

A further critical point is the setup of the hybrid system, i.e. if the computers are co-located or at different places. Some applications require low-latency, including algorithms with feedback-loops and small execution times, and therefore need a tight connection between QC and HPC system, hence, a co-location. A vision that is far from being implemented is to have both units on the same chip. Yet, it is also unclear how much one would gain compared to the co-location approach. Moreover, data security implementation and maintenance will be facilitated by in case of co-location. For applications where latency is not critical, such as runtimes beyond the network’s latency, the systems can be located at places that are beneficial for their performance and energy consumption. For example an underground location of a shields the system from cosmic radiation, whereas locating an HPC system close to a renewable power sources and/or in a cold climate, decreases the system’s environmental footprint and/or its total energy consumption (i.e. costs).

Within the project High Performance Computer – Quantum Simulator hybrid (HPCQS) two quantum simulators will be fully integrated into co-located HPC environments. One system will be installed at the Grand Équipement National de Calcul Intensif, the other system at JSC providing a HPC-QCS hybrid. Within the project a software fleet will be developed, which manages code compilation, execution and data workflow between the systems. Moreover, the hybrid system will be made available to European researchers by the end of the project via the Jülich UNified Infrastructure for Quantum computing (JUNIQ), which provides access to various quantum devices. Since this is one of the first projects worldwide integrating QC and HPC, it will lift HPC to the next level and provide lessons learned with respect to software architecture, quantum hardware design, and latency implications. According to integrating the quantum simulator into JSC’s HPC environment, all quantum devices available via JUNIQ will be integrated, if possible.