Brown, Kenneth RDalvi, Aniket Sudeep2025-07-022025-07-022025https://hdl.handle.net/10161/32678<p>Quantum computing has introduced a new paradigm in computing that has the potential to exceed the capabilities of any classical computing system. Unlike classical computing which stores information in bits, quantum computing encodes exponentially more information in qubits. It uses quantum mechanical phenomena to store and manipulate this information to accelerate the execution of a certain class of problems. This new model of computation requires us to rethink how to interface with such a computer, necessitating innovations in computer science, computer engineering, and electrical engineering. In this dissertation, I present my research on software, compilation, and architecture for quantum systems with a view towards scalability and optimization.</p><p>Starting at the bottom of the software stack, I identify a repeated compilation overhead at the control software/hardware level which can take up 5%-80% of the total program execution time. I address this by developing a device-level partial-compilation (DLPC) technique that reduces compilation overhead to nearly constant through parametrization of gate/pulse parameters of quantum subroutines. Moving higher up in the stack, I propose pulselib - a graph-based pulse-level representation. This scalable, device-agnostic, parameter-based, lightweight representation allows for maximum retention of high-level pulse information before the pulse instruction is realized. It enables seamless access at the pulse level through semantics for scheduling, phase synchronization, and variability, while also lending itself to optimizations and transformation. At the gate level of the software stack, I propose a real-time noise-aware qubit mapping protocol. I identify that current qubit mapping protocols rely on the last known device calibration data, which may be stale at the time of circuit compilation. To this end, I propose the use of scalable benchmarking protocols to be run as part of the compilation pipeline. This would provide sufficient information about the noise landscape of the device which can be used for noise-aware qubit mapping. Preliminary results show that circuits executed on the mapping chosen by our technique achieve up to a 20.4% improvement over the current state-of-the-art approach of using a heuristic based on device calibration data. Lastly, I describe my work on building a modular software architecture for classical simulation of quantum circuits. It enables users to identically run a quantum circuit on hardware and in simulation using any simulator backend of their choosing. This serves to be foundational work towards building a virtual twin of a quantum computer which could be used as a test-bed for testing applications and informing device calibration routines.</p><p>Through this dissertation, I have identified and addressed some sub-optimalities in various layers of the quantum software stack. However, as the field progresses towards an era of fault-tolerant quantum computation, it opens up more room for innovation and optimization in compilation and software architectures. I conclude this dissertation with a discussion of a few such ideas.</p>https://creativecommons.org/licenses/by-nc-nd/4.0/Computer engineeringComputer scienceQuantum physicsControl SoftwareQuantum ArchitectureQuantum CompilationQuantum ComputingQuantum SimulationSoftware EngineeringDissertation