WAGONN: Weight Bit Agglomeration in Crossbar Arrays for Reduced Impact of Interconnect Resistance on DNN Inference Accuracy

TL;DR

SWANN is a weight shuffling technique for crossbar array DNN accelerators that significantly improves inference accuracy under high interconnect resistance with minimal hardware overhead.

Contribution

The paper introduces SWANN, a novel weight shuffling method that mitigates interconnect resistance effects in crossbar arrays for DNN inference, enhancing accuracy and efficiency.

Findings

SWANN improves ResNet-20/CIFAR-10 accuracy from 47.78% to 83.5%.

SWANN can be combined with Partial-Word-LineActivation for further gains.

SWANN adds less than 1% energy overhead and about 16% area overhead.

Abstract

Deep neural network (DNN) accelerators employing crossbar arrays capable of in-memory computing (IMC) are highly promising for neural computing platforms. However, in deeply scaled technologies, interconnect resistance severely impairs IMC robustness, leading to a drop in the system accuracy. To address this problem, we propose SWANN - a technique based on shuffling weights in crossbar arrays which alleviates the detrimental effect of wire resistance on IMC. For 8T-SRAM-based 128x128 crossbar arrays in 7nm technology, SWANN enhances the accuracy from 47.78% to 83.5% for ResNet-20/CIFAR-10. We also show that SWANN can be used synergistically with Partial-Word-LineActivation, further boosting the accuracy. Moreover, we evaluate the implications of SWANN for compact ferroelectric-transistorbased crossbar arrays. SWANN incurs minimal hardware overhead, with less than a 1% increase in energy consumption. Additionally, the latency and area overheads of SWANN are ~1% and ~16%, respectively when 1 ADC is utilized per crossbar array.