RAISE: MemComputing: From Software to Hardware

University Of California-San Diego, La Jolla CA

Investigators

Abstract

MemComputing is an unconventional, physics-based computing paradigm that exploits the ubiquitous time non-locality of any physical system, namely memory of their past dynamics. Although it is fundamentally different from the traditional computing model, and even quantum computing, it can be emulated in software on modern computers. This is because it relies on non-quantum dynamical systems to compute. The software emulation of the digital version of MemComputing machines (DMMs) has already shown great advantages compared to traditional algorithms in the solution of some hard computational problems. However, there are many academic and industrial applications for which a hardware implementation of these machines is not only preferable, but even necessary. The main goal of this RAISE project is precisely to implement MemComputing in hardware, and test its efficiency on a representative set of computationally-intensive problems. This project will then set the foundations towards the ultimate goal of realizing such DMMs as co-processors in our traditional computing systems to both enhance and greatly extend their functions. State-of-the-art field-programmable gate arrays (FPGAs) will be used to implement these computing machines. To assist their hardware design, a theory of MemComputing maps will be developed to analyze the dynamics of DMMs, understand their modus operandi, and suggest improvements. Starting from small-size prototypes, medium- and large-size DMMs will be implemented in FPGAs. Time, space, and energy scalability in the solution of various combinatorial optimization problems will be used as metrics to improve on their design and push the bounds imposed by the hardware (such as the amount of logic cells in FPGAs). The project goes to the very core of what it means to compute, with far-reaching implications. In fact, its interdisciplinarity will help the professional development of a future work force skilled in successfully navigating across traditional disciplinary boundaries, as it is becoming ever more relevant in both industry and academia. Finally, MemComputing is that type of transformative technology that may provide a clear competitive edge to its users in the rapidly evolving scientific and industrial landscape of today. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.