Research and Development of System Software for Quantum-HPC Hybrid Computing

Most supercomputers and quantum computers employ batch job systems to manage jobs from many different users, based on operation policies designed to maximize the throughput of each system. To maximize throughput when the dependencies between quantum jobs and HPC jobs are relatively simple, we will develop workflow engines that submit jobs to supercomputers and quantum computers in an appropriate order based on the dependencies between the jobs, making efficient use of quantum computers and supercomputers across both systems.

We are developing the RPC offloading library for (2), together with a quantum computing request scheduler to control the priority of quantum computing requests coming from a supercomputer and those coming directly from quantum computer users, whether issued by users themselves or through the workflow of (1).

The quantum computing request scheduler works in front of the quantum computing backend server, and all requests to quantum computers on our platform must pass through the quantum computing request scheduler. The scheduler prioritizes requests from supercomputers over those issued by individual quantum computer users, so as to avoid leaving supercomputers idle while a request waits in the queue. This high priority for quantum computing requests from a supercomputer is granted only for quantum-HPC hybrid jobs in which the quantum and HPC tasks are tightly coupled.

The quantum computing request scheduler will also support user authorization, confirming that a request from a supercomputer comes from a legitimate user who holds usage rights for both the supercomputer and the quantum computer.

The communication libraries and interfaces needed for a supercomputer to communicate with the quantum computing request scheduler in order to request work from a quantum computer will also be developed in this project.

Another important mission of this project is porting quantum software development kits (SDKs) to supercomputers. Quantum SDKs such as Qiskit, Qibo, and cuQuantum provide various functions, including transpilers, circuit optimization, and error correction and mitigation for the results of quantum circuit execution, in addition to supporting the execution of quantum circuits on quantum computers. By porting these functions to supercomputers, we will enable existing quantum computer users to make easy use of the quantum-HPC hybrid platform. To this end, we will extend the RPC offloading software to enable integration with quantum SDKs.

Furthermore, as the number of qubits in quantum computers increases, these SDK functions are expected to require greater computational resources. By making use of supercomputers, we will make large-scale quantum circuit optimization, quantum circuit transpilation, and error correction and mitigation possible for future large-scale quantum computers and quantum circuits.