A Low Power In-Memory Multiplication andAccumulation Array with Modified Radix-4 Inputand Canonical Signed Digit Weights

TL;DR

This paper introduces a low power in-memory computing scheme using modified Radix-4 and Canonical Signed Digit techniques, significantly reducing power consumption and improving efficiency for AI tasks.

Contribution

It proposes a novel digital CIM scheme with modified radix-4 and CSD weights, fully utilizing digital memory properties to enhance reliability and power efficiency.

Findings

Reduces 1x1 ratio by over 78% on LeNet and AlexNet

Improves computing efficiency by 41.6% on average

Achieves 60.68 TOPS/s/W at 8-bit fixed-point

Abstract

A mass of data transfer between the processing and storage units has been the leading bottleneck in modern Von-Neuman computing systems, especially when used for Artificial Intelligence (AI) tasks. Computing-in-Memory (CIM) has shown great potential to reduce both latency and power consumption. However, the conventional analog CIM schemes are suffering from reliability issues, which may significantly degenerate the accuracy of the computation. Recently, CIM schemes with digitized input data and weights have been proposed for high reliable computing. However, the properties of the digital memory and input data are not fully utilized. This paper presents a novel low power CIM scheme to further reduce the power consumption by using a Modified Radix-4 (M-RD4) booth algorithm at the input and a Modified Canonical Signed Digit (M-CSD) for the network weights. The simulation results show that M-Rd4 and M-CSD reduce the ratio of $1\times1$ by 78.5\% on LeNet and 80.2\% on AlexNet, and improve the computing efficiency by 41.6\% in average. The computing-power rate at the fixed-point 8-bit is 60.68 TOPS/s/W.