This comprehensive analysis of a newly developed asynchronous circuit family covers circuit theory, practical circuits, design tools and an example of the design of a simple asynchronous microprocessor using the circuit family.

Formal verification means having a mathematical model of a system, a language for specifying desired properties of the system in a concise, comprehensible and unambiguous way, and a method of proof to verify that the specified properties are satisfied. When the method of proof is carried out substantially by machine, we speak of automatic verification. Symbolic Model Checking deals with methods of automatic verification as applied to computer hardware. The practical motivation for study in this area is the high and increasing cost of correcting design errors in VLSI technologies. There is a growing demand for design methodologies that can yield correct designs on the first fabrication run. Moreover, design errors that are discovered before fabrication can also be quite costly, in terms of engineering effort required to correct the error, and the resulting impact on development schedules. Aside from pure cost considerations, there is also a need on the theoretical side to provide a sound mathematical basis for the design of computer systems, especially in areas that have received little theoretical attention.

CMOS Processors and Memories addresses the-state-of-the-art in integrated circuit design in the context of emerging computing systems. New design opportunities in memories and processor are discussed. Emerging materials that can take system performance beyond standard CMOS, like carbon nanotubes, graphene, ferroelectrics and tunnel junctions are explored. CMOS Processors and Memories is divided into two parts: processors and memories. In the first part we start with high performance, low power processor design, followed by a chapter on multi-core processing. They both represent state-of-the-art concepts in current computing industry. The third chapter deals with asynchronous design that still carries lots of promise for future computing needs. At the end we present a “hardware design space exploration” methodology for implementing and analyzing the hardware for the Bayesian inference framework. This particular methodology involves: analyzing the computational cost and exploring candidate hardware components, proposing various custom architectures using both traditional CMOS and hybrid nanotechnology CMOL. The first part concludes with hybrid CMOS-Nano architectures. The second, memory part covers state-of-the-art SRAM, DRAM, and flash memories as well as emerging device concepts. Semiconductor memory is a good example of the full custom design that applies various analog and logic circuits to utilize the memory cell’s device physics. Critical physical effects that include tunneling, hot electron injection, charge trapping (Flash memory) are discussed in detail. Emerging memories like FRAM, PRAM and ReRAM that depend on magnetization, electron spin alignment, ferroelectric effect, built-in potential well, quantum effects, and thermal melting are also described. CMOS Processors and Memories is a must for anyone serious about circuit design for future computing technologies. The book is written by top notch international experts in industry and academia. It can be used in graduate course curriculum.

Shared memory multiprocessors are becoming the dominant architecture for small-scale parallel computation. This book is the first to provide a coherent review of current research in shared memory multiprocessing in the United States and Japan. It focuses particularly on scalable architecture that will be able to support hundreds of microprocessors as well as on efficient and economical ways of connecting these fast microprocessors. The 20 contributions are divided into sections covering the experience to date with multiprocessors, cache coherency, software systems, and examples of scalable shared memory multiprocessors.

1.1 Background There are many paradigmatic statements in the literature claiming that this is the decade of parallel computation. A great deal of research is being de voted to developing architectures and algorithms for parallel machines with thousands, or even millions, of processors. Such massively parallel computers have been made feasible by advances in VLSI (very large scale integration) technology. In fact, a number of computers having over one thousand pro cessors are commercially available. Furthermore, it is reasonable to expect that as VLSI technology continues to improve, massively parallel computers will become increasingly affordable and common. However, despite the significant progress made in the field, many funda mental issues still remain unresolved. One of the most significant of these is the issue of a general purpose parallel architecture. There is currently a huge variety of parallel architectures that are either being built or proposed. The problem is whether a single parallel computer can perform efficiently on all computing applications.

In the summer of 1981 I was asked to consider the possibility of manufacturing a 600,000 transistor microprocessor in 1985. It was clear that the technology would only be capable of manufacturing 100,000-200,000 transistor chips with acceptable yields. The control store ROM occupied approximately half of the chip area, so I considered adding spare rows and columns to increase ROM yield. Laser-programmed polysilicon fuses would be used to switch between good and bad circuits. Since only half the chip area would have redundancy, I was concerned that the increase in yield would not outweigh the increased costs of testing and redundancy programming. The fabrication technology did not yet exist, so I was unable to experimentally verify the benefits of redundancy. When the technology did become available, it would be too late in the development schedule to spend time running test chips. The yield analysis had to be done analytically or by simulation. Analytic yield analysis techniques did not offer sufficient accuracy for dealing with complex structures. The simulation techniques then available were very labor-intensive and seemed more suitable for redundant memories and other very regular structures [Stapper 80J. I wanted a simulator that would allow me to evaluate the yield of arbitrary redundant layouts, hence I termed such a simulator a layout or yield simulator. Since I was unable to convince anyone to build such a simulator for me, I embarked on the research myself.