A Review of SRAM-based Compute-in-Memory Circuits

TL;DR

This paper reviews SRAM-based compute-in-memory circuits, comparing digital and analog approaches, analyzing their architectures, advantages, challenges, and emerging hybrid solutions for efficient in-memory computing.

Contribution

It provides a comprehensive tutorial and comparison of SRAM-based CIM architectures, highlighting recent developments and hybrid approaches.

Findings

DCIM offers high precision and good scalability.

ACIM provides power and area efficiency for medium-precision tasks.

Hybrid CIM architectures combine strengths of both DCIM and ACIM.

Abstract

This paper presents a tutorial and review of SRAM-based Compute-in-Memory (CIM) circuits, with a focus on both Digital CIM (DCIM) and Analog CIM (ACIM) implementations. We explore the fundamental concepts, architectures, and operational principles of CIM technology. The review compares DCIM and ACIM approaches, examining their respective advantages and challenges. DCIM offers high computational precision and process scaling benefits, while ACIM provides superior power and area efficiency, particularly for medium-precision applications. We analyze various ACIM implementations, including current-based, time-based, and charge-based approaches, with a detailed look at charge-based ACIMs. The paper also discusses emerging hybrid CIM architectures that combine DCIM and ACIM to leverage the strengths of both approaches.