While architectural and algorithmic innovations have prolonged the viability of conventional semiconductor devices, diminishing returns are eventually inevitable. Emerging technologies often present alternative and more promising paths forward, but their adoption is not without challenges. Developing advanced fabrication processes is both costly and slow, and many existing circuit, digital logic, and microarchitecture solutions are, at best, suboptimal for these technologies due to fundamental differences in device physics.
This dissertation advocates for superconductor electronics, promising for applications ranging from classical computing acceleration to integrated quantum and sensing systems, and introduces practical co-optimizations that exploit the technology's strengths while circumnavigating its limitations. The objective is to examine and exploit the opportunities created by the absence of resistance and the intrinsic characteristics of Josephson junctions at various levels of the computing stack, rather than blindly force-fitting established abstractions.
To this end, the presented research focuses on four aspects critical to the development of a computer system: logic, design automation, fanout, and memory. In each case, the overarching goal is to maximize the utility of every Josephson junction. For the first three aspects, this means optimizing logic density, while for memory, the focus is on demonstrating how leveraging inexpensive data movement can improve both circuit complexity and energy efficiency.