scientific article; zbMATH DE number 194430
PAX Computer is a book about the hardware, the software, and the programming methods of one particular parallel computer, namely the Japanese PAX computer, which was built by Prof. Tsutomo Hoshino and his colleagues. The PAX computer was primarily constructed for computations involved in nuclear power research; but it is in fact a general-purpose computer for scientific computations and for modeling and simulations of a variety of physical phenomena. The book was originally published in Japanese in 1985. Except for the addition of a few new references (co-authored by the author of the book), the English translation of the book has not been updated. The book consists of eight chapters which range from technical descriptions of the computer to very general comments about parallel programming methods. Chapter 1 surveys the development of data-parallel computers until approx. 1984, with emphasis on the ILLIAC IV. This chapter also includes a discussion of Prof. Hoshino's main thesis: modern-type computers should be simple and specialized to the computations that they are intended to carry out (in this instance: scientific computations). Chapter 2 discusses the three typical types of models which are used in scientific computations: continuous, lumped-parameter, and particle systems. Next, the concepts of `autonomy' and `proximity' are introduced, and it is pointed out how these concepts reappear in the natural phenomena, in their physical models, and in the numerical computations - thus indeed making parallel processing possible. This chapter is very general in nature and can be read by anyone interested in parallel programming. In contrast to this, Chapters 3-5 are specific for the PAX computer. Based on the discussion in Chapter 2, Chapter 3 provides the arguments why PAX was constructed as a MIMD-type parallel computer with global synchronization (which is unusual). Chapter 4 gives details of the hardware and software of PAX, and Chapter 5 essentially contains material about data communication of the PAX computer. A lot of this material is more for computer designers than for general programmers and users. Chapter 6 (which constitutes 40 \% of the book) gives thorough discussions of 11(!) examples of the use of PAX: formulation, parallel implementation, and efficiency. The chapter is mainly intended as a `proof' of the usefulness of the PAX computer; but it also serves as a good illustration of how to actually formulate and program parallel implementations of physical systems. The material in Chapter 7 is intended as general remarks about parallel programming methods; but unfortunately only a few examples are given. Finally, Chapter 8 discusses the future of massive parallel computers. A lot of the material in this book may be quite interesting for computer designers - in particular, it is interesting to see a description of a MIMD-type computer with global synchronization. Unfortunately, since the material dates back from 1984-85, much of this material is quite outdated now in view of the many commercially available data-parallel computers. The only part of the book which can still be read today with general interest to a broad range of scientists are Chapters 2 and 6 about physical models and their parallel implementation.
This page was built for publication:
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4692385)