Summary
Spiking neural networks are a type of artificial brain used in smart devices like cameras, but running them efficiently can be hard. The researchers designed FlexSpIM, a chip that stores and processes data together in a flexible way to save power and speed up calculations. This chip can adjust how detailed its calculations are and uses clever tricks to reuse information, making it faster and less energy-hungry. Tests showed that FlexSpIM works well on real gesture recognition tasks, using less energy and time than older designs. This could help make smarter, quicker, and more efficient devices for edge applications.
compute-in-memoryspiking neural networksdigital acceleratoroperand resolutiondataflowenergy efficiencyinference latencylayer-wise stationarityedge computingDVS gesture dataset
Authors
Nicolas Chauvaux, Adrian Kneip, Charlotte Frenkel
Abstract
Compute-in-memory (CIM) accelerators for spiking neural networks (SNNs) offer a promising solution for achieving $μ$s-level inference latency and ultra-low energy in edge vision applications. However, their limited flexibility at both circuit and system levels restricts their deployment across diverse workloads. This work introduces FlexSpIM, a digital CIM architecture supporting arbitrary operand resolution and shape within a unified storage for weights and neuron states (i.e., membrane potentials). These circuit-level capabilities enable a layer-level hybrid weight- and output-stationary dataflow, maximizing operand reuse and reducing costly on- and off-chip data movement during SNN execution. Measurement results from a fabricated FlexSpIM prototype in 40-nm CMOS demonstrate competitive 1-bit-normalized energy efficiency and higher throughput compared with prior fixed-precision digital CIM-based SNN accelerators, while providing bitwise resolution reconfiguration. Evaluated on the IBM DVS gesture dataset, FlexSpIM achieves 95.8% accuracy while enabling up to 45% energy and 52% latency reductions in large-scale systems compared with fixed stationarity approaches.