Unicorn: A Multicore Neuromorphic Processor with Flexible Fan-In and Unconstrained Fan-Out for Neurons
TimeThursday, July 14th11:15am - 11:37am PDT
Location3005, Level 3
Event Type
Research Manuscript
AI/ML Design: Circuits and Architecture
DescriptionNeuromorphic processor is popular due to high energy efficiency. However, the evolution of spiking neural networks (SNNs) is challenging to traditional neuromorphic architectures, that is, the difficulties of mapping and executing SNNs with large neuron fan-in and fan-out. This paper proposes Unicorn, a multicore digital neuromorphic processor with unconstrained fan-out and reconfigurable fan-in of neurons via spike train sliding multicasting mechanism and neuron units combination method. Experimental results show that Unicorn can support larger neuron fan-out than the baselines and thus gain xx\% performance improvement and xx\% energy efficiency improvement for it eliminates the process of multiple mapping and executing.