DSLR-CNN: Efficient CNN Acceleration using Digit-Serial Left-to-Right Arithmetic

TL;DR

This paper introduces DSLR-CNN, an efficient CNN accelerator using digit-serial left-to-right arithmetic, achieving significant improvements in performance, latency, and energy efficiency over existing designs.

Contribution

It presents a novel left-to-right digit-serial arithmetic approach for CNN acceleration, enabling pipelined processing and substantial performance gains.

Findings

Peak performance up to 569.11x higher than existing designs

Energy efficiency improved by up to 44.75x

Significant reductions in latency and power consumption

Abstract

Digit-serial arithmetic has emerged as a viable approach for designing hardware accelerators, reducing interconnections, area utilization, and power consumption. However, conventional methods suffer from performance and latency issues. To address these challenges, we propose an accelerator design using left-to-right (LR) arithmetic, which performs computations in a most-significant digit first (MSDF) manner, enabling digit-level pipelining. This leads to substantial performance improvements and reduced latency. The processing engine is designed for convolutional neural networks (CNNs), which includes low-latency LR multipliers and adders for digit-level parallelism. The proposed DSLR-CNN is implemented in Verilog and synthesized with Synopsys design compiler using GSCL 45nm technology, the DSLR-CNN accelerator was evaluated on AlexNet, VGG-16, and ResNet-18 networks. Results show significant improvements across key performance evaluation metrics, including response time, peak performance, power consumption, operational intensity, area efficiency, and energy efficiency. The peak performance measured in GOPS of the proposed design is 4.37x to 569.11x higher than contemporary designs, and it achieved 3.58x to 44.75x higher peak energy efficiency (TOPS/W), outperforming conventional bit-serial designs.