PRECISE's Safe Autonomy Seminar: Hardware-Software Interface for Neuromorphic Computers

Neuromorphic computers are emerging computing systems that operate on the principles of the central nervous system. They implement neurons and synapses in hardware, supporting biology-inspired synaptic plasticity. These systems can perform several different types of scientific computations with significantly lower energy footprints compared to a conventional CPU-based computer. Future high-performance neuromorphic computers are expected to aggregate multiple heterogeneous neuromorphic hardware nodes to solve scientific computations that are far too complex for single-node hardware. Despite the progress made on the hardware and technology fronts, the software stacks for these computers have remained largely unexplored. As a result, even a small neuromorphic computer, where neurons and synapses are in the millions, requires an enormous amount of time and expertise to program and visualize results. This complex interface is expected to become a programming bottleneck for systems that can have several orders more neurons and synapses than today’s systems. In this talk I will show our recent works on hardware-software interface for such systems. I will discuss a compiler tool chain that we have developed in our lab to translate a user's machine learning program to low-level languages that can be interpreted by neuromorphic systems. I will also present interesting resource optimization strategies to improve program performance, energy, reliability, and security. Finally, I will present an Operating System like framework that we have developed to allow programmers to easily deploy their machine learning programs on neuromorphic computers as well as to perform full-stack co-design of neuromorphic systems.