PASS: Exploiting Post-Activation Sparsity in Streaming Architectures for CNN Acceleration

TL;DR

This paper introduces a method to leverage post-activation sparsity in streaming CNN accelerators, significantly improving efficiency over existing instruction-based sparse accelerators.

Contribution

It presents a novel approach to exploit post-activation sparsity specifically in streaming CNN architectures, addressing previous challenges and demonstrating substantial efficiency gains.

Findings

Achieves 1.41x to 1.93x higher efficiency (GOP/s/DSP)

Addresses challenges in exploiting sparsity in streaming architectures

Demonstrates effectiveness on modern CNN benchmarks

Abstract

With the ever-growing popularity of Artificial Intelligence, there is an increasing demand for more performant and efficient underlying hardware. Convolutional Neural Networks (CNN) are a workload of particular importance, which achieve high accuracy in computer vision applications. Inside CNNs, a significant number of the post-activation values are zero, resulting in many redundant computations. Recent works have explored this post-activation sparsity on instruction-based CNN accelerators but not on streaming CNN accelerators, despite the fact that streaming architectures are considered the leading design methodology in terms of performance. In this paper, we highlight the challenges associated with exploiting post-activation sparsity for performance gains in streaming CNN accelerators, and demonstrate our approach to address them. Using a set of modern CNN benchmarks, our streaming sparse accelerators achieve 1.41x to 1.93x efficiency (GOP/s/DSP) compared to state-of-the-art instruction-based sparse accelerators.