Efficient Reprogramming of Memristive Crossbars for DNNs: Weight Sorting and Bit Stucking

TL;DR

This paper presents a new method to significantly reduce memristor reprogramming frequency in DNN crossbars by organizing weights and selectively reprogramming memristors, thereby extending device endurance without sacrificing accuracy.

Contribution

It introduces weight sorting and selective memristor reprogramming techniques to minimize reprogramming frequency in memristive crossbars for DNNs.

Findings

Reprogramming reduced by 3.7x for ResNet-50

Reprogramming reduced by 21x for ViT-Base

Model accuracy maintained within 1% margin

Abstract

We introduce a novel approach to reduce the number of times required for reprogramming memristors on bit-sliced compute-in-memory crossbars for deep neural networks (DNNs). Our idea addresses the limited non-volatile memory endurance, which restrict the number of times they can be reprogrammed. To reduce reprogramming demands, we employ two techniques: (1) we organize weights into sorted sections to schedule reprogramming of similar crossbars, maximizing memristor state reuse, and (2) we reprogram only a fraction of randomly selected memristors in low-order columns, leveraging their bit-level distribution and recognizing their relatively small impact on model accuracy. We evaluate our approach for state-of-the-art models on the ImageNet-1K dataset. We demonstrate a substantial reduction in crossbar reprogramming by 3.7x for ResNet-50 and 21x for ViT-Base, while maintaining model accuracy within a 1% margin.