System Design with Memristor Technologies
Memristors are a new class of circuit element with the ability to change their resistance value while retaining memory of their current and past resistances. Their small form factor, high density, and fast switching times have sparked research into their applications in modern memory hierarchies. However, these new components pose system design challenges, as well as opportunities. System Design with Memristor Technologies explores design solutions for memristors, covering research and development trends in memristor technology, fabrication, modelling, and applications, and the design and implementation of arithmetic units using memristors. The book begins with an introduction to the principles of system design with memristors, then goes on to address memristor logic gates, arithmetic units for adders, multipliers and dividers, and improved and optimised adder, multiplier and divider designs. The final chapters draw conclusions from the topics covered and explore potential future trends in research into system designs with memristor technologies. This book is essential reading for research scientists and electronics engineers interested in the use of memristors in future system architectures, specifically focused on the areas of arithmetic units, non-Von-Neumann architectures, and logic-in-memory
Inspec keywords: dividing circuits; memristor circuits; logic design; adders; multiplying circuits; logic gates
Other keywords: memristor-based adder designs; memristor-based divider designs; system design; memristor logic gates; adder arithmetic units; multiplier arithmetic units; memristor-based multiplier designs; divider arithmetic units; memristor technologies
Subjects: Logic and switching circuits; Logic elements; General electrical engineering topics; General and management topics; Digital circuit design, modelling and testing; Logic design methods; Logic circuits
- Book DOI: 10.1049/PBCS038E
- Chapter DOI: 10.1049/PBCS038E
- ISBN: 9781785615610
- e-ISBN: 9781785615627
- Page count: 366
- Format: PDF
-
Front Matter
- + Show details - Hide details
-
p.
(1)
-
1 Introduction
- + Show details - Hide details
-
p.
1
–16
(16)
This paper presents the introduction of memristors. Memristors were first hypothesized by Chua in 1971. Memristors, or memristive devices, are nonvolatile circuit components that have a nonlinear relationship between electric charge and magnetic flux. These devices were hypothesized as the missing fourth circuit element, joining the long-standing resistor, inductor, and capacitor. This text presents prior work in this field, the proposed arithmetic designs, and results and suggestions for future work.
-
2 Memristor logic gates
- + Show details - Hide details
-
p.
17
–59
(43)
There are numerous approaches to implementing logic with memristors. The most popularized approach is the IMPLY operation, most commonly known from HP Labs' groundbreaking study. However, since this study, numerous other approaches have been proposed, which offer differing properties and trade-offs. Some of these approaches include memristors-as-drivers (MAD) gates, hybridCMOS gates, MAGIC gates, threshold gates, Zhang, and others. This chapter will explore two of these approaches in detail: the IMPLY and MAD operations because they have the widest applicability and the most research in their use. The remaining approaches will be discussed briefly for completion.
-
3 Adder arithmetic units
- + Show details - Hide details
-
p.
61
–71
(11)
Adders are a universal component of nearly every computing system, whether traditional or modern. Even the most fundamental machine, which performs no arithmetic, requires an adder to calculate address offsets and branch instructions. Adders are also generally less complex than other arithmetic units found in system. This chapter introduces four different adders: a ripple-carry adder, a carry-look-ahead adder, a carry-select adder, and a conditional-sum adder. The traditional CMOS circuits for each of the adders are described, and the performance and area of these designs are each analyzed. The adders vary in terms of their complexity. The simplest design is the ripple-carry adder, and the most complex design is the carry-look-ahead adder. In general, as the complexity of the adder increases, the speed and performance improve.
-
4 Multiplier arithmetic units
- + Show details - Hide details
-
p.
73
–77
(5)
Multipliers are becoming ubiquitous in modern computing systems. With the drive for big data and machine-learning applications, multipliers have become an underlying construct for nearly all state-of-the-art workloads. Depending on the specific system, design constraints, and expected application, the optimal multiplier implementation can vary. Sometimes, a simpler, slower multiplier is preferred and sometimes a more complex, faster multiplier is. This chapter presents four popular multipliers, which span these two metrics: shift-and-add multipliers, Booth multipliers, array multipliers, and Dadda multipliers. The performance and area of these designs are each analyzed in the CMOS domain.
-
5 Divider arithmetic units
- + Show details - Hide details
-
p.
79
–86
(8)
Dividers are commonly found in floating point units. These arithmetic units are not only found in standard computing systems but also in high-performance computing as well. Similar to the presented adders and multipliers in the previous two chapters, design constraints, system characteristics, and expected applications all affect the selection of the optimal divider implementation. Sometimes, a simpler, slower divider is preferred and sometimes a more complex, faster divider is. This chapter presents three dividers that represent varying complexity and speed: binary restoring dividers; Sweeney, Robertson, and Tocher (SRT) dividers; and Goldschmidt dividers. The CMOS design for each of these dividers is explained and analyzed in terms of speed and area.
-
6 Memristor-based adder designs
- + Show details - Hide details
-
p.
87
–191
(105)
This chapter presents various designs of memristor-based adders. IMPLY, hybridCMOS, threshold gate, and MAD approaches are employed to implement ripplecarry, carry-select, conditional-sum, and carry-lookahead adders. Each design is described and analyzed in terms of complexity and delay. Completed schematics are also given.
-
7 Memristor-based multiplier designs
- + Show details - Hide details
-
p.
193
–263
(71)
This chapter presents contributions in the design and implementation of memristor based multipliers. Minimal prior work exists for memristor-based designs for multipliers. The complexity of these designs coupled with the complexity of memristor models and their programming leads to a high design overhead. However, there have been a few prior works that have explored multipliers in the context of memristors for the IMPLY approach. Similar to Chapter 6 on adders, the focus is on four different multipliers: shiftand-add, Booth, array, and Dadda multipliers. For each multiplier, implementations using IMPLY, hybrid-CMOS, threshold gate, and MAD approaches are examined. Each implementation is explained and analyzed in terms of complexity and delay. Due to the increasing complexity of these designs, the CSTG threshold gate implementations are not considered. The component area of CSTG implementations precludes them from being desirable for these units. Recall that a single threeinput CSTG threshold gate requires three memristors, three resistors, and ten MOSFETs. Thus, for some of the more complex designs, only the GOTO pair implementations are presented. Complete schematics and simulations are also given.
-
8 Memristor-based divider designs
- + Show details - Hide details
-
p.
265
–326
(62)
This chapter presents completed work in the design and implementation of memristor-based dividers. IMPLY, hybrid-CMOS, threshold-gate, and MAD approaches to binary-restoring, SRT, and Goldschmidt dividers are described and analyzed in terms of complexity and delay. Completed schematics and simulations are also given.
-
9 Proposed future work
- + Show details - Hide details
-
p.
327
–337
(11)
Given the infancy of memristor research, there are many avenues for future directions in this area. It is essential that students, researchers, and industry colleagues continue to actively explore the world of memristors and its impact on the future of modern systems.
-
10 Conclusion
- + Show details - Hide details
-
p.
339
–340
(2)
This text aims to provide a thorough and complete introduction to memristive devices and their use in modern arithmetic units. The memristive devices can exhibit a wide range of characteristics and properties depending on their chemical and physical makeup. Often times, the expected application for the device determines many of these characteristics. The flexibility of the design of memristive devices is just one of its strongpoints as a candidate for future system design. Memristors are nonvolatile devices making them prime candidates for use in memory structures. Specifically, crossbars have been shown as the frontrunner in memristor-based storage. Memristors are capable of exhibiting high density and low power in these contexts. They have competitive read and write times and endurance. Perhaps most importantly, memristors are compatible with traditional CMOS designs as well as 3D architectures.
-
Back Matter
- + Show details - Hide details
-
p.
(1)