Abstract
In modern von Neumann systems, the data are stored in the memory but are processed in a separate unit. Data transfer between these units incurs energy and delay, which are several orders of magnitude greater than the energy and delay incurred by the computation itself. Recent works on in-memory computing (IMC) have been proposed to reduce the overhead of data transfer. Further reduction became possible with the emergence of memristive memory technologies. Numerous logic-in-memory techniques with memristors have been proposed where the computation is performed using the memory arrays. In this chapter, we overview different memristor-based logic techniques. As a case study of memristor-based logic in an IMC system, we demonstrate how Memristor Aided loGIC (MAGIC) allows NOR gates to be performed within a memristive crossbar array structure. We describe a potential memristive Memory Processing Unit (mMPU) where MAGIC NOR is employed as the basis for all data processings. We then show how to perform different image processing tasks within the mMPU to obtain superior performance and energy efficiency for these tasks over other state-of-the-art memristive logic systems.
Original language | English |
---|---|
Title of host publication | Memristive Devices for Brain-Inspired Computing |
Subtitle of host publication | From Materials, Devices, and Circuits to Applications - Computational Memory, Deep Learning, and Spiking Neural Networks |
Publisher | Elsevier |
Pages | 175-194 |
Number of pages | 20 |
ISBN (Electronic) | 9780081027820 |
DOIs | |
State | Published - 1 Jan 2020 |
Externally published | Yes |
Keywords
- Algorithms
- In-memory computing
- MAGIC
- Memristors
- Von Neumann bottleneck
ASJC Scopus subject areas
- Engineering (all)