ADRA: Extending Digital Computing-in-Memory with Asymmetric Dual-Row-Activation

TL;DR

This paper introduces ADRA, a novel digital computing-in-memory technique using asymmetric dual-row-activation, enabling efficient computation of a broader class of functions including non-commutative operations like subtraction, with significant energy-delay improvements.

Contribution

It proposes an asymmetric wordline biasing scheme that allows single-cycle memory read and computation of all Boolean functions, extending CiM capabilities to non-commutative operations.

Findings

Enables simultaneous single-cycle memory read and Boolean function computation.

Supports computation of non-commutative functions like subtraction.

Achieves 23.2% - 72.6% reduction in energy-delay product compared to standard methods.

Abstract

Computing in-memory (CiM) has emerged as an attractive technique to mitigate the von-Neumann bottleneck. Current digital CiM approaches for in-memory operands are based on multi-wordline assertion for computing bit-wise Boolean functions and arithmetic functions such as addition. However, most of these techniques, due to the many-to-one mapping of input vectors to bitline voltages, are limited to CiM of commutative functions, leaving out an important class of computations such as subtraction. In this paper, we propose a CiM approach, which solves the mapping problem through an asymmetric wordline biasing scheme, enabling (a) simultaneous single-cycle memory read and CiM of primitive Boolean functions (b) computation of any Boolean function and (c) CiM of non-commutative functions such as subtraction and comparison. While the proposed technique is technology-agnostic, we show its utility for ferroelectric transistor (FeFET)-based non-volatile memory. Compared to the standard near-memory methods (which require two full memory accesses per operation), we show that our method can achieve a full scale two-operand digital CiM using just one memory access, leading to a 23.2% - 72.6% decrease in energy-delay product (EDP).