Four Simulators of the DANNA Neuromorphic Computing Architecture

Abstract

DANNA is a computing architecture, designed in 2014 to meld features of recurrent, spiking, plastic neuromorphic computing systems with very efficient hardware implementations. Its hardware design and FPGA implementation preceded any software support or simulation. Therefore, the development of simulators for DANNA had multiple, and sometimes conflicting goals. In this paper, we describe four separate simulators for DANNA, including their motivations and details of their implementations. We present a performance evaluation, and conclude with lessons learned. We intend this work to be useful to researchers involved in the co-design of novel hardware and software systems.