Practical Parallel Computing

Practical Parallel Computing provides information pertinent to the fundamental aspects of high-performance parallel processing. This book discusses the development of parallel applications on a variety of equipment.

Organized into three parts encompassing 12 chapters, this book begins with an overview of the technology trends that converge to favor massively parallel hardware over traditional mainframes and vector machines. This text then gives a tutorial introduction to parallel hardware architectures. Other chapters provide worked-out examples of programs using several parallel languages. This book deals as well with benchmarking and performance estimation on parallel machines. The final chapter provides a structured, flexible methodology for selecting a parallel machine and for integrating it into operations.

This book is a valuable resource for readers who are confronted with the practical realities of parallel computing for the first time. Mid-level technical managers, algorithm designers, computer scientists, and doctorate-level mathematicians will also find this book extremely useful.

Preface

Acknowledgments

Part 1 Preliminaries

1 The Status and Future of Massively Parallel Processing

1.1 Technology Trends Favoring Parallel Architectures

1.2 Obstacles Inhibiting Commercial Success

1.3 Trends and Considerations

Summary of Chapter 1

2 Can Parallel Machines Be Used Efficiently?

2.1 A Parallel Parable: Building a Wall

2.2 Amdahl's Law

2.3 Examples

Summary of Chapter 2

3 An Introduction to Hardware Architectures

3.1 Impacts of Hardware Architecture

3.2 A Primer on Interconnection Networks

3.3 SIMD Machines

3.4 Distributed Memory MIMD Machines

3.5 Shared Memory MIMD Machines

Summary of Chapter 3

Part 2 Software Issues

4 Shared Memory Parallel Language Constructs

4.1 Basic Concepts for Shared Memory Parallel Programming

4.2 The Sample Problem on an SGI Challenge SMP

4.3 Observations and Considerations

Summary of Chapter 4

5 Message Passing

5.1 An Overview of the Message-Passing Library

5.2 The Sample Problem on an nCube 2

5.3 Express

5.4 Intel Paragon

5.5 Observations and Considerations

Summary of Chapter 5

6 SIMD and Array-Based Languages

6.1 An Overview

6.2 MPL on the MasPar MP-1

6.3 C* on the Connection Machine

Summary of Chapter 6

7 Linda

7.1 A Linda Primer

7.2 The Sample Problem in Linda

7.3 Observations and Considerations

Summary of Chapter 7

8 The Development Environment for Parallel Software

8.1 Compilers

8.2 Debugging Parallel Code

8.3 Profilers and Load Balancing

8.4 Other Tools

Summary of Chapter 8

9 Operating System Issues

9.1 Multiple Users

9.2 Virtual Address Spaces

9.3 Scheduling

9.4 Virtual Processors

9.5 I/O

9.6 Open Systems

Summary of Chapter 9

Part 3 Management Issues

10 Benchmarking Parallel Applications

10.1 Dependence on Problem Size and Machine Size

10.2 Publicly Available Parallel Benchmarks

10.3 Scaling Sequential Performance to Parallel Performance

10.4 Estimating Performance

10.5 Questions to Ask Vendors

Summary of Chapter 10

11 Porting and Developing Parallel Applications

11.1 Porting Strategies

11.2 Developing Parallel Applications

11.3 Examples

Summary of Chapter 11

12 Matching Applications to Architectures

12.1 A Methodology

12.2 Integrating a Parallel Machine into Existing Operations

Summary of Chapter 12

Appendix A The Sample Problem

Listing for Sample Problem

Appendix B SGI Challenge

Appendix C nCube

Appendix D Express

Appendix E Intel Paragon

Appendix F MasParMP-1

Appendix G C* on the Connection Machine

Appendix H Linda

Appendix I Two Recent Machines

1.1 The SP-Series from IBM

Node Architecture

Interprocessor Communications

I/O Software

1.2 Exemplar Series from Convex Corp.

System Architecture

System Software

1.3 Summary and Comparison

References

Index