Performance evaluation of acceleration of convolutional layers on OpenEdgeCGRA

TL;DR

This paper evaluates the performance of convolutional layer acceleration on OpenEdgeCGRA, demonstrating that direct convolution with weight parallelism offers superior latency and energy efficiency over CPU implementations.

Contribution

It presents an analysis of different convolution mapping strategies on OpenEdgeCGRA, highlighting the optimal approach for latency and energy efficiency.

Findings

Direct convolution with weight parallelism outperforms CPU in latency and energy.

OpenEdgeCGRA achieves 3.4x energy and 9.9x latency improvements.

Transforming convolution into matrix multiplication is less efficient on this hardware.

Abstract

Recently, efficiently deploying deep learning solutions on the edge has received increasing attention. New platforms are emerging to support the increasing demand for flexibility and high performance. In this work, we explore the efficient mapping of convolutional layers on an open-hardware, low-power Coarse-Grain Reconfigurable Array (CGRA), namely OpenEdgeCGRA. We explore both direct implementations of convolution and solutions that transform it into a matrix multiplication through an Im2col transformation, and experiment with various tensor parallelism axes. We show that for this hardware target, direct convolution, coupled with weight parallelism reaches the best latency and energy efficiency, outperforming a CPU implementation by 3.4x and 9.9x in terms of energy and latency, respectively.