High Performance Computing and the Art of

High Perfomance Computing and the Art of Parallel Computingprovides a non-technical introduction to High Performance Computing applications together with advice about how beginners can start to write parallel programs. Using case studies to show where HPC has already been used and providing examples for other areas of application to geography, GIS and the social sciences, it provides a plain and practical introduction to a subject that is usually surrounded by complex computer jargon.

List of figures

ix

List of tables

x

List of appendices

xii

Dedication and acknowledgements

xiii

High-performance computing: why bother with it?

1

(11)

An HPC point of view

1

(1)

HPC stimulates new research and creates new research opportunities

2

(1)

Why parallel processing is important

3

(2)

Parallel computing is the future of HPC

5

(2)

Aims and objectives

7

(1)

Fostering a computational culture

8

(2)

Plan of the book

10

(2)

High-performance computing applications in geography and GIS

12

(35)

Introduction

12

(3)

Parallel programming

15

(2)

Geocomputation

17

(2)

Raising HPC awareness

19

(7)

HPC applications in geography and GIS

26

(3)

Some examples of HPC applications in geography

29

(13)

Parallel GIS applications

42

(3)

Overcoming access barriers

45

(2)

Parallel and high-performance computing: concepts, principles and theory

47

(27)

What is parallel computing?

47

(5)

Why parallel processing is important

52

(5)

Highly and massively parallel processing

57

(5)

Examples of thinking in parallel

62

(2)

Can paralle machines ever be used efficiently?

64

(5)

Building a wall in a parallel way

69

(2)

A brief history of parallel computing

71

(2)

Conclusions

73

(1)

Types of parallel-processing hardware and programming paradigms

74

(20)

Automatic parallelisation software

74

(2)

Computer architectures

76

(6)

The three principal types of HPC hardware

82

(6)

Levels of parallelism and identifying them

88

(2)

Programming models

90

(1)

Examples of each type of parallelism

91

(1)

Conclusions

92

(2)

Programming vector supercomputers

94

(45)

Introduction

94

(1)

The secrets of vector processing

95

(4)

Vectorising your code

99

(1)

Optimisation of performance

100

(4)

A case study in vector processing using the Mark 1 geographical analysis machine as an example

104

(15)

Case study 2: orgin-constrained spatial interaction model

119

(3)

Conclusions

122

(17)

Shared-loop and data parallel programming

139

(39)

Introduction

139

(1)

Multi-tasking on shared-memory MIMD machines

140

(7)

Parallelisation strategies

147

(4)

Data parallel programming

151

(4)

Conclusions

155

(23)

Parallel programming using simple message passing

178

(33)

Introduction

178

(2)

Message passing?

180

(2)

Message-passing software

182

(2)

How to use MPI for SPMD

184

(2)

Example 1: probably the world's simplest MPI program that could be useful

186

(2)

Example 2: sum M numbers

188

(8)

Example 3: a data parallel spatial interaction in MPI

196

(7)

Conclusions

203

(8)

Parallelising the geographical analysis machine using MPI

211

(24)

A data parallel GAM using MPI

211

(2)

Where is the parallelism in the GAM?

213

(2)

Loop 4 message passing

215

(2)

Some alternative message-passing schemes

217

(2)

Doing even better by task farming

219

(2)

A task-farming GAM

221

(1)

Improvements?

222

(2)

Loop 0 complications

224

(1)

Multiple task farms

225

(1)

More advanced MPI routines

225

(1)

Conclusions

225

(10)

Optimising performance and debugging hints

235

(16)

Introduction

235

(1)

First optimise your algorithm rather than fiddling with code

236

(2)

Now start to fiddle

238

(1)

Scaleable performance

238

(2)

Exploit Amdahl's law

240

(1)

Some MPI optimisation secrets

241

(2)

Debugging parallel code is harder than serial code

243

(1)

Debugging message passing

244

(3)

Defensive coding

247

(1)

Shared-memory debugging

248

(1)

Message-passing debugging

248

(2)

Conclusions

250

(1)

Putting it all together

251

(17)

Background

251

(1)

Introduction to benchmarking

251

(2)

The spatial interaction model as a benchmark code

253

(3)

The high-performance Fortran version

256

(2)

The message-passing code using MPI

258

(1)

The bulk synchronous parallel model

259

(1)

Measuring performance using MPI and serial code

259

(2)

A comparison of HPF and MPI codes

261

(4)

Conclusions

265

(3)

Epilogue for geographers and social scientists

268

(9)

The global challenge

268

(4)

What has HPC got to do with any of this?

272

(1)

Revising the definition of geocomputation

272

(2)

A GIS--HPC research agenda

274

(2)

HPC futures in geography, etc.

276

(1)

References and further reading

277

(6)

Index

283

What is included with this book?

The New copy of this book will include any supplemental materials advertised. Please check the title of the book to determine if it should include any access cards, study guides, lab manuals, CDs, etc.

The Used, Rental and eBook copies of this book are not guaranteed to include any supplemental materials. Typically, only the book itself is included. This is true even if the title states it includes any access cards, study guides, lab manuals, CDs, etc.

FREE SHIPPING

Amazon no longer offers textbook rentals. We do!

High Performance Computing and the Art of Parallel Programming: An Introduction for Geographers, Social Scientists and Engineers

Summary

Table of Contents

Supplemental Materials