Parallel Supercomputing in Single Instruction Multiple Data Architectures

The aim of this work is to provide a comprehensive review of Single-Instruction-Multiple-Data machines, a type of parallel processing computer that has grown to importance in recent years. It describes various aspects of this technology, including the architectural concept, its history, hardware implementations, major programming languages, algorithmic methods, representative applications and an assessment of benefits and drawbacks. There are data on neural nets, seismic analysis, electromagnetic scattering, network flow optimization and orbit analysis. Focusing on the data parallel paradigm instead of the more familiar program parallel scheme, the book is organized into nine chapters. The first two provide an introduction and background information covering basic concepts and a description of early SIMD computers. The text then addresses specific machines from the first SIMD supercomputer (Iliac IV) through several contemporary designs to some example research computers. The final chapter contains a commentary on the lessons learned.