Flow-Based Computing on Memristor Crossbars and its application in Computer Vision for Edge Detection

Date

2024

Authors

Pannu, Jodh Singh

Journal Title

Journal ISSN

Volume Title

Publisher

Abstract

The end of Dennard scaling and the approaching saturation of transistor density for memory devices has resulted in the need for alternate methods of computation. We explore the fabrication, design and use of in-memory computation architecture, namely, memristor nanoscale crossbars. Memristors are non-volatile two terminal devices. They are essentially programmable variable resistors that retain the programmed resistance values even after the device loses its power. This property is used to create memristor crossbar circuits that can perform in-memory computation on the data stored in these crossbars. Flow-based computing has been developed to make use of the unique architecture of memristor crossbars.

In this thesis we answer the following questions:

How can flow-based computing be used to implement boolean formulas on memristor cross-bars.

How do we synthesis memristor crossbar circuits for more specialised computation problems, for example edge detection in images.

Providing a proof of concept by fabricating real memristor crossbar circuits.

Description

Keywords

Computer Vision, Emerging Architecture, Flow-based Computing, Memristor

Citation

Department

Computer Science