L2R-CIPU: Efficient CNN Computation with Left-to-Right Composite Inner Product Units

TL;DR

This paper introduces L2R-CIPU, a novel composite inner-product unit based on left-to-right arithmetic, significantly accelerating CNN computations and improving efficiency metrics on hardware implementations.

Contribution

The paper presents a new hardware unit for CNN acceleration that outperforms prior architectures in speed and efficiency using a left-to-right composite inner product approach.

Findings

Achieves 1.06x to 6.22x performance improvement

Provides 4.8x to 15x higher TOPS/W efficiency

Delivers 4.51x to 53.45x increased TOPS/mm2 density

Abstract

This paper proposes a composite inner-product computation unit based on left-to-right (LR) arithmetic for the acceleration of convolution neural networks (CNN) on hardware. The efficacy of the proposed L2R-CIPU method has been shown on the VGG-16 network, and assessment is done on various performance metrics. The L2R-CIPU design achieves 1.06x to 6.22x greater performance, 4.8x to 15x more TOPS/W, and 4.51x to 53.45x higher TOPS/mm2 than prior architectures.