Browse Books

Reviewer: John A. Fulcher

Rarely does one encounter a new computer science text that upon first reading shows every sign of becoming the standard reference work in its field. Equally rarely does a book live up to the publishers prepublicity. It is a pleasure to report that both statements are true of Patterson and Hennessys work. During the past 12 months, I have been evaluating books for a new intermediate-level computer architecture course at the University of Wollongong. To date I have examined more than 20 prospective texts. At the end of the day, I had no hesitation in adopting Patterson and Hennessy for this course. What makes this book stand out to such an extent from the competition First and foremost, it is written by two eminent researchers in the field. Not only did they carry out fundamental research into reduced instruction set computers (RISC), they followed this through into commercially viable products (MIPS and Sun SPARC). Nevertheless, and contrary to what one might expect, this book does not unduly emphasize the RISC approach. Second, the principles introduced throughout the text are related back to real-world, commercially successful computers (in particular the IBM 360370, DEC VAX, and Intel i80x86). I definitely have the impression that the authors are speaking from experience more often than not, particularly in the last sections of each chapter, “Fallacies and Pitfalls” and “Historical Perspective.” It is refreshing to read a book that dispels commonly held misconceptions and places “new” ideas in a historical context; many such ideas are not so new at all. Third, the numerous worked examples clearly illustrate and amplify the principles just introduced. The “Definitions, Trivia, Formulas, and Rules of Thumb” guide on the inside front cover is an excellent feature. Fourth, the book has already passed through extensive class testing and peer review. As a consequence, it appears to be relatively bug-free and robust. Moreover, both an instructors manual and accompanying UNIX support software are available from Morgan Kaufmann; the software includes benchmarks, cache traces, cache and instruction set simulators, and a compiler. The book is divided into ten chapters: Fundamentals of Computer Design Performance and Cost Instruction Set Design: Alternatives and Principles Instruction Set Examples and Measurements of Use Basic Processor Implementation Techniques Pipelining Vector Processors Memory-Hierarchy Design InputOutput Future Directions I can find little to fault in this book. I would have preferred to see more on systolic arrays and digital signal processors (DSP), which are covered in Section 10.6, “Special Purpose Processors,” possibly at the expense of the appendices on System360, VAX, and i8086. I would also have liked to see the chapter on parallel processing expanded. I had recently thought of writing my own book on computer architecture, prompted by the lack of suitable books then available. In the wake of Hennessy and Patterson, however, I would have to ask myself, “Why bother ”— t he definitive text has just been written. The authors may have left room for a new book on parallel architectures, but probably not for another four or five years; this may reflect the relative infancy of this area. I had no reservations in adopting this book for our computer architecture course. Moreover, I am sure many other universities worldwide will adopt what will no doubt become the classic reference on computer architecture. Let us hope the publisher does the right thing for students and releases a paperback version shortly.