E-raamat: Algorithm Engineering for Integral and Dynamic Problems

Algorithm engineering aims to design the most cost-efficient computational machine that will execute an algorithm given a set of constraints, such as minimal performance or the availability of technology.
This book addresses algorithm engineering in a parallel setting, that of regular array processor. It focuses on powerful engineering techniques - regular array syntheses - which simultaneously embody best practice and are well-founded in mathematical theories. The core of regular synthesis is the expression of an algorithm in terms of its computations' data dependencies and their subsequent and systematic transformation onto regular array processors.
A criticism that is often levelled against regular array synthesis is that it can only be applied to restricted classes of algorithms. In this book, after a review of the basic principles of regular array synthesis, we will show how regular array synthesis can be extended to include classes of algorithms traditionally thought to be beyond its domain of application.

Provides a comprehensive treatment of algorithm transformations for the derivation of regular processor arrays. Rapanotti (Open University) develops the characterization of classes of integral and dynamic algorithms, and the provision of techniques for their systematic treatment within the framework of established synthesis methods. The basic idea is to transform the initial algorithm specification into a specification with data dependencies of increased regularity so that corresponding arrays can be obtained by a direct application of the standard mapping techniques. Four case studies illustrate the application of the techniques. Annotation c. Book News, Inc., Portland, OR (booknews.com)

Algorithm engineering allows computer engineers to produce a computational machine that will execute an algorithm as efficiently and cost-effectively as possible given a set of constraints, such as minimal performance or the availability of technology. Addressing algorithm engineering in a parallel setting, regular array syntheses offer powerful computation and embody best practice, but often face the criticism that they are applicable only to restricted classes of algorithms.

Algorithm Engineering for Integral and Dynamic Problems reviews the basic principles of regular array synthesis and shows how to extend its use into classes of algorithms traditionally viewed to be beyond its domain of application. The author discusses the transformation of the initial algorithm specification into a specification with data dependencies of increased regularity in order to obtain corresponding regular arrays by direct application of the standard mapping techniques. The book includes a review of the basic principles of regular array synthesis followed by applications of these techniques to well-known algorithms, concluding with numerous case studies to illustrate the methods.

Researchers and practitioners in algorithm engineering will find that this text significantly extends their understanding of the applications of regular array synthesis and regular array processors beyond the traditionally narrow field of relevance.