Special Purpose Computers (eBook)

Berni J. Alder (Herausgeber)

eBook Download: PDF

2014 | 1. Auflage
294 Seiten
Elsevier Science (Verlag)
978-1-4832-6698-5 (ISBN)

Infos

Special Purpose Computers describes special-purpose computers and compares them to general-purpose computers in terms of speed and cost. Examples of computers that were designed for the efficient solution of long established algorithms are given, including Navier-Stokes hydrodynamic solvers, classical molecular dynamic machines, and Ising model computers. Comprised of seven chapters, this volume begins by documenting the progress of the CalTech Concurrent Computation Program and its evolution from computational high-energy physics to a supercomputer initiative, with emphasis on the lessons learned including computer architecture issues and the trade-offs between in-house and commercial development. The reader is then introduced to the QCD Machine, a special-purpose parallel supercomputer that was designed and built to solve the lattice quantum chromodynamics problem. Subsequent chapters focus on the Geometry-Defining Processors and their application to the solution of partial differential equations; the Navier-Stokes computer; parallel processing using the Loosely Coupled Array of Processors (LCAP) system; and the Delft Ising system processor. The design and implementation of the Delft molecular-dynamics processor are also described. This book will be of interest to computer engineers and designers.

Inhalt

Front Cover 1
Special Purpose Computers 4
Copyright Page 5
Table of Contents 6
Contributors 8
Introduction 10
Chapter 1. The Hypercube and the Caltech Concurrent Computation Program: A Microcosm of Parallel Computing 12
I. Original Motivation and Historical Overview 12
II. Hypercube Hardware 20
III. Domain Decomposition and the Hypercube as a General-Purpose Supercomputer 30
IV. Software for the Hypercube 43
V. The Future 48
References 48
Chapter 2. The QCD Machine 52
I. Introduction 52
II. Background 52
III. Architecture 55
IV. Programming 62
V. Implementation 65
VI. Experience 74
References 75
Chapter 3. Geometry —Defining Processors for Partial Differential Equations 78
I. Introduction 78
II. The Finite-Element Method: Geometry and Parallelism 79
III. Geometry-Defining Processors 83
IV. Geometry Input and Mesh Generation 86
V. Parallel Solution of Partial Differential Equations 88
VI. Prototype Implementation of Mesh-Generation GDPs 93
VII. Performance Estimates for Computational GDPs 100
Acknowledgements 106
References 106
Chapter 4. A Navier— Stokes Computer 108
I. Introduction 108
II. Architectural Overview of NSC 115
III. NSC Hardware Organization 118
IV. NSC Software 127
References 145
Chapter 5. Parallel Processing with the Loosely Coupled Array of Processors System 146
I. Introduction 146
II. Initial Configuration for LCAP 148
III. System Considerations 152
IV. Parallel Programming Strategy and Considerations 155
V. Initial Applications and Performance 159
VI. Extended LCAP, Applications and Performance 178
VII. Summary 191
References 191
Chapter 6. Design and Implementation of the Delft Molecular-Dynamics Processor 194
I. Introduction 194
IL Computational Aspects of Molecular Dynamics 196
III. Design Considerations for a Special-Purpose Molecular-Dynamics Processor 202
IV. Design of the MD Processor 208
V. Implementation 232
VI. Software 238
VII. Conclusions 241
References 242
Acknowledgments 243
Chapter 7. The Delft Ising System Processor 244
I. Introduction 244
II. Design Characteristics 245
III. Hardware Architecture 247
IV. The Random-Number Generator 261
V. Performance 271
VI. Software 272
VII. Summary of Results 274
VIII. Current Activities 288
Acknowledgments 290
References 290
Index 292

Produkt-Details

• Erscheint lt. Verlag: 10.5.2014
• Sprache: englisch
• Themenwelt: Mathematik / Informatik ► Informatik ► Theorie / Studium
• Themenwelt: Informatik ► Weitere Themen ► Hardware
• ISBN-10: 1-4832-6698-2 / 1483266982
• ISBN-13: 978-1-4832-6698-5 / 9781483266985