Browse Books

Reviewer: Dorothy Bollman

The emphasis in this book is on the design of parallel algorithms, using a class of basic communication operations as building blocks. The first four chapters present the necessary background: Introduction Models of Parallel Computers Basic Communication Operators Performance and Scalability of Parallel Systems Subsequent chapters are devoted to specific classes of algorithms: (5) Dense Matrix Algorithms (6) Sorting (7) Graph Algorithms (8) Search Algorithms for Discrete Optimization P roblems (9) Dynamic Programming (10) Fast Fourier Transform (11) Solving Sparse Systems of Linear Equations Chapter 12 deals with “Systolic Algorithms and their Mapping onto Parallel Computers.” Finally, chapter 13, “Parallel Programming,” treats some specifics of programming on several current architectures, including the nCUBE 2, iPSC860, and CM5, and describes several languages used for parallel programming, including C*, CM Fortran, and Fortran D. Here the authors might have mentioned the emerging role of functional programming languages, such as Sisal, Id, and Haskell, in parallel computing. Each chapter concludes with bibliographic remarks, a list of problems, and a list of references. The book concludes with an appendix that summarizes the complexity functions and order analysis notation used, as well as both author and topic indices. The topic index contains references, for the most part, to first mentions of topics. For example, if one looks up “PRAM,” one is referred to the general description of the PRAM model, rather than to the many algorithm analyses in terms of PRAMs. Several distinctive features of the book make it attractive as both a text and a reference. First, it is organized around algorithms rather than architectures. This approach makes it more practical to use the book as a text for a course in parallel algorithms. One could cover the first four chapters and a selection of topics from the remaining chapters, depending on the interests of the audience. All of the material on sorting, for example, is concentrated in one chapter. This approach contrasts with other books on parallel algorithms that are organized around architectures, rather than around algorithms, in which one might find material on sorting scattered throughout sections dealing with arrays of processors, trees of processors, or hypercubes. Second, the coverage of algorithms is unusually broad. The inclusion of such topics as discrete optimization and dynamic programming makes the book appealing to a wider audience, including those interested in applications of parallel computing such as grand challenge problems. Third, the inclusion of scalability as a performance metric for parallel algorithms is especially welcome. Unfortunately, other authors only allude to this important concept, if they even mention it. This book fulfills the authors objectives as a text for an upper-level undergraduate or beginning graduate course or as a useful reference for parallel algorithm designers and application programmers. The exposition is clear and illustrated with an abundance of pictures and examples. References are extensive and current.