IMPULSE: A 65-nm Digital Compute-in-Memory Macro With Fused Weights and Membrane Potential for Spike-Based Sequential Learning Tasks
Amogh Agrawal, Mustafa Ali, Minsuk Koo, Nitin Rathi, Akhilesh Jaiswal, Kaushik Roy
Abstract
The inherent dynamics of the neuron membrane potential in spiking neural networks (SNNs) allows the processing of sequential learning tasks, avoiding the complexity of recurrent neural networks. The highly sparse spike-based computations in such spatiotemporal data can be leveraged for energy efficiency. However, the membrane potential incurs additional memory access bottlenecks in current SNN hardware. To that effect, we propose a 10T-SRAM compute-in-memory (CIM) macro, specifically designed for state-of-the-art SNN inference. It consists of a fused weight ( W <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> MEM</sub> ) and membrane potential ( V <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> MEM</sub> ) memory and inherently exploits sparsity in input spikes leading to ~97.4% reduction in energy-delay product (EDP) at 85% sparsity (typical of SNNs considered in this work) compared to the case of no sparsity. We propose staggered data mapping and reconfigurable peripherals for handling different bit precision requirements of W <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> MEM</sub> and V <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> MEM</sub> , while supporting multiple neuron functionalities. The proposed macro was fabricated in 65-nm CMOS technology, achieving energy efficiency of 0.99 TOPS/W at 0.85-V supply and 200-MHz frequency for signed 11-bit operations. We evaluate the SNN for sentiment classification from the IMDB dataset of movie reviews and achieve within ~1% accuracy difference and ~ 5× higher energy efficiency compared to a corresponding long short-term memory network.