Hardware/software Co-design for Neuromorphic Systems

Abstract

Transistor technology for electronic computer systems is now at the single digit nanometer scale. This enormous advance through sustained efforts over more than sixty years has resulted in computers that are extremely efficient in terms of the energy per unit of computation. This progress in hardware was arguably driven by the demand for computation, as software systems and digital technology became integrated with more and more of our lives. Despite this progress in device technology, general-purpose microprocessors-the heart of a modern computer-can still be viewed as a “von Neumann” computer with control, storage, arithmetic, and input/output devices. As the demand for computation grows unabated while the scaling of transistor technology slows down, alternate approaches to further reducing the energy per unit of computation are required.