Softwaredevelopment - Middleware

The quantum computing technology stack must span the entire spectrum from hardware to applications. To use a quantum algorithm to solve an application problem, in addition to appropriate application-specific frameworks and high-level or programming languages, other software tools must be developed and made available that operate at the compiler level. The goal is to translate the abstract programming language into a language that can be understood by the hardware that controls and reads out the qubits. There is significant potential for optimization in this translation process; for example, the number of “expensive” quantum operations can be reduced through circuit optimization, and hardware characteristics can be taken into account to minimize interference. Overall, this offers the opportunity to make the best possible use of the available quantum hardware.

At Fraunhofer IIS, building on expertise in the fields of AI and mathematical optimization, methods for circuit optimization and so-called qubit routing were developed. Furthermore, classical and quantum-based methods for problem reduction were developed, which effectively improve the capabilities of current quantum hardware. The corresponding quantum-based method (known as “circuit cutting”) enables distributed or parallelized quantum computing and represents a software-based approach to scaling quantum computing. Ongoing projects for the development of quantum computing middleware at Fraunhofer IIS include the Munich Quantum Valley (MQV) consortium QACI (Quantum Algorithms for Application, Cloud & Industry), the MQV flagship project KID-QC^2 (AI-supported design for scalable, efficient, and highly structured quantum circuits for quantum chemistry), and the BMWK-funded project QuaST (quantum-enabling services and tools for industrial applications).

At Fraunhofer ITWM, a standardized compiler for quantum hardware is being developed as part of the QCStack project. Also worth mentioning is the Quantum Architecture Search from the section on quantum machine learning, as it takes into account hardware properties such as the number of qubits, interconnections, and error rates, ensuring that transpiling is directly incorporated into the design of the quantum circuit.