PREFACE ix

The book concerns itself with general purpose computers which are intended to be widely available commercially. The engineering of computers for highly specialized applications, for which only a few copies are built, is not treated. More over, because not all major principles of computer architecture and computer engineering are embodied in the DEC computers, the reader may want to examine other designs, as well. For example, the reader cannot learn about descriptor architectures, array processors, list-processing machines, or general purpose emulators from this book.

At one time consideration was given to postponing the publication of a book until 1982, at which time DEC will celebrate its twenty-fifth anniversary. This idea was rejected because another five years would further impede the collection of data about the early machines. More importantly, the twenty-year period of DEC modules and computers (1957-1977) has extended from the early second generation to the fourth generation. Today, the processor of several DEC computers occupies a single large-scale integrated circuit consisting of several thou sand transistors, whereas in 1957 only one transistor could be fabricated on a single piece of germanium. In another five years, the design, manufacture, and distribution of computers will be radically different - so much so as to merit a new book.

We expect an increasingly larger number of people to be involved in computer engineering and hence students of this material, because we expect computers as we know them today will disappear within ten years! With the processor-on-a- chip, the number of computer systems designers (users) has risen by several orders of magnitude.

In the area of large computer systems, the buyers and users are also clearly the computer designers: they select components (from the set of available components) and interconnect them to form specific structures. It is essential for us all to have a model of the price, performance, and reliability parameters and how they vary with time. Previous generations have focused first on the invention of the computer, next on the understanding of price/performance tradeoffs, and most recently on manufacturing - especially the fabrication of the semiconductors that now drive computer evolution. In the next five years, design will focus on applications: conventional applications will be more efficient, computers will be extended to reach new applications, and life-cycle costs will receive more attention. For the computer engineer, the evolution of DEC machines provides an excellent perspective on the influence of applications on design. For those of us who must deal with design goals, constraints, and objective functions to improve reliability, availability and maintainability, it is imperative that we first clearly understand previous design problems.

For the programmers who use computers and are a part of the computer design process, understanding this material is mandatory in order to know the rules of the game. We say comparatively little about software, other than how it has influenced hardware design. The increasing role of software functions in the hard ware domain is a clear process that has allowed (and forced) computer architecture to change. The engineering of DEC software will be treated in subsequent