Analysis, transformation and optimization for high perfomance parallel computing

Abstract

This book studies hardware and software specifications at algorithmic level from the point of measuring and extracting the potential parallelism hidden in them. It investigates the possibilities of using this parallelism for the synthesis and optimization of highperformance software and hardware implementations. The basic single-block flow model is a result of the algorithm transformation, and is a basis for developing efficient methods of synthesis and optimization of parallel implementations. It supports the generation and optimization of computational pipelines and concurrent net algorithms, which lead to higher performances of the computing systems. This book is useful in training of scientific researchers and graduate students. It is also useful for teaching students and undergraduates in computer science at universities.