S2Engine: A Novel Systolic Architecture for Sparse Convolutional Neural Networks

TL;DR

S2Engine is a new systolic architecture designed to fully exploit CNN sparsity, significantly improving speed and energy efficiency by transmitting compressed data and enabling dynamic data selection in convolution.

Contribution

It introduces a novel systolic architecture that leverages CNN sparsity through compressed data transmission and dynamic data selection, overcoming previous limitations.

Findings

Achieves 3.2x speedup over naive systolic array.

Attains 3.0x energy efficiency improvement.

Effectively exploits CNN sparsity with compressed data flow.

Abstract

Convolutional neural networks (CNNs) have achieved great success in performing cognitive tasks. However, execution of CNNs requires a large amount of computing resources and generates heavy memory traffic, which imposes a severe challenge on computing system design. Through optimizing parallel executions and data reuse in convolution, systolic architecture demonstrates great advantages in accelerating CNN computations. However, regular internal data transmission path in traditional systolic architecture prevents the systolic architecture from completely leveraging the benefits introduced by neural network sparsity. Deployment of fine-grained sparsity on the existing systolic architectures is greatly hindered by the incurred computational overheads. In this work, we propose S2Engine $-$ a novel systolic architecture that can fully exploit the sparsity in CNNs with maximized data reuse. S2Engine transmits compressed data internally and allows each processing element to dynamically select an aligned data from the compressed dataflow in convolution. Compared to the naive systolic array, S2Engine achieves about $3.2\times$ and about $3.0\times$ improvements on speed and energy efficiency, respectively.