采用教程形式，从简短的编程实例起步，一步步编写更有挑战性的程序。重点介绍分布式内存和共享式内存的程序设计、调试和性能评估。使用MPI、PTrlread和OperIMP等编程模型，强调实际动手开发并行程序。

并行编程已不仅仅是面向专业技术人员的一门学科。如果想要全面开发机群和多核处理器的计算能力，那么学习分布式内存和共享式内存的并行编程技术是不可或缺的。由Peter S.Pacheco编著的《并行程序设计导论(英文版)》循序渐进地展示了如何利用MPI、PThread和OperlMP开发高效的并行程序，教给读者如何开发、调试分布式内存和共享式内存的程序，以及对程序进行性能评估。

Contents

CHAPTER 1 Why Parallel Computing?

1.1 Why We Need Ever-Increasing Performance

1.2 Why We're Building Parallel Systems

1.3 Why We Need to Write Parallel Programs

1.4 How Do We Write Parallel Programs?

1.5 What We'll Be Doing

1.6 Concurrent, Parallel, Distributed

1.7 The Rest of the Book

1.8 A Word of Warning

1.9 Typographical Conventions

1.10 Summary

1.11 Exercises

CHAPTER 2 Parallel Hardware and Parallel Software

2.1 Some Background

 2.1.1 The von Neumann architecture

 2.1.2 Processes, multitasking, and threads

2.2 Modifications to the von Neumann Model

 2.2.1 The basics of caching

 2.2.2 Cache mappings

 2.2.3 Caches and programs: an example

 2.2.4 Virtual memory

 2.2.5 Instruction-level parallelism

 2.2.6 Hardware multithreading.

2.3 Parallel Hardware

 2.3.1 SIMD systems

 2.3.2 MIMD systems 32

 2.3.3 Interconnection networks

 2.3.4 Cache coherence

 2.3.5 Shared-memory versus distributed-memory

2.4 Parallel Software 47

 2.4.1 Caveats 47

 2.4.2 Coordinating the processes/threads

 2.4.3 Shared-memory 49

 2.4.4 Distributed-memory

 2.4.5 Programming hybrid systems

2.5 Input and Output 56

2.6 Performance 58

 2.6.1 Speedup and efficiency

 2.6.2 Amdahl's law 61

 2.6.3 Scalability

 2.6.4 Taking timings

2.7 Parallel Program Design

 2.7.1 An example

2.8 Writing and Running Parallel Programs

2.9 Assumptions

2.10 Summary

 2.10.1 Serial systems

 2.10.2 Parallel hardware

 2.10.3 Parallel software

 2.10.4 Input and output

 2.10.5 Performance.

 2.10.6 Parallel program design

 2.10.7 Assumptions

2.11 Exercises

CHAPTER 3 Distributed-Memory Programming with MPI

3.1 Getting Started 84

 3.1.1 Compilation and execution

 3.1.2 MPI programs

 3.1.3 MPI Init and MPI Finalize

 3.1.4 Communicators, MPI Comm size and MPI Comm rank

 3.1.5 SPMD programs

 3.1.6 Communication

 3.1.7 MPI Send

 3.1.8 MPI Recv

 3.1.9 Message matching

 3.1.10 The status p argument 92

 3.1.11 Semantics of MPI Send and MPI Recv 93

 3.1.12 Some potential pitfalls 94

3.2 The Trapezoidal Rule in MPI 94

 3.2.1 The trapezoidal rule 94

 3.2.2 Parallelizing the trapezoidal rule 96

Contents xiii

3.3 Dealing with I/O 97

 3.3.1 Output 97

 3.3.2 Input100

3.4 Collective Communication 101

 3.4.1 Tree-structured communication 102

 3.4.2 MPI Reduce 103

 3.4.3 Collective vspoint-to-point communications 105

 3.4.4 MPI Allreduce 106

 3.4.5 Broadcast 106

 3.4.6 Data distributions 109

 3.4.7 Scatter 110

 3.4.8 Gather 112

 3.4.9 Allgather 113

3.5 MPI Derived Datatypes 116

3.6 Performance Evaluation of MPI Programs 119

 3.6.1 Taking timings 119

 3.6.2 Results 122

 3.6.3 Speedup and efficiency 125

 3.6.4 Scalability 126

3.7 A Parallel Sorting Algorithm 127

 3.7.1 Some simple serial sorting algorithms 127

 3.7.2 Parallel odd-even transposition sort 129

 3.7.3 Safety in MPI programs 132

 3.7.4 Final details of parallel odd-even sort 134

3.8 Summary 136

3.9 Exercises 140

3.10 Programming Assignments .147

CHAPTER 4 Shared-Memory Programming with Pthreads .151

4.1 Processes, Threads, and Pthreads 151

4.2 Hello, World 153

 4.2.1 Execution 153

 4.2.2 Preliminaries 155

 4.2.3 Starting the threads 156

 4.2.4 Running the threads 157

 4.2.5 Stopping the threads 158

 4.2.6 Error checking 158

 4.2.7 Other approaches to thread startup159

4.3 Matrix-Vector Multiplication 159

4.4 Critical Sections 162

xiv Contents

4.5 Busy-Waiting 165

4.6 Mutexes .168

4.7 Producer-Consumer Synchronization and Semaphores 171

4.8 Barriers and Condition Variables 176

 4.8.1 Busy-waiting and a mutex 177

 4.8.2 Semaphores 177

 4.8.3 Condition variables 179

 4.8.4 Pthreads barriers 181

4.9 Read-Write Locks 181

 4.9.1 Linked list functions 181

 4.9.2 A multi-threaded linked list 183

 4.9.3 Pthreads read-write locks 187

 4.9.4 Performance of the various implementations 188

 4.9.5 Implementing read-write locks 190

4.10 Caches, Cache Coherence, and False Sharing 190

4.11 Thread-Safety 195

 4.11.1 Incorrect programs can produce correct output 198

4.12 Summary 198

4.13 Exercises 200

4.14 Programming Assignments .206

CHAPTER 5 Shared-Memory Programming with OpenMP .209

5.1 Getting Started 210

 5.1.1 Compiling and running OpenMP programs 211

 5.1.2 The program 212

 5.1.3 Error checking215

5.2 The Trapezoidal Rule 216

 5.2.1 A first OpenMP version 216

5.3 Scope of Variables 220

5.4 The Reduction Clause .221

5.5 The parallel for Directive 224

 5.5.1 Caveats 225

 5.5.2 Data dependences 227

 5.5.3 Finding loop-carried dependences 228

 5.5.4 Estimating 229

 5.5.5 More on scope231

5.6 More About Loops in OpenMP: Sorting .232

 5.6.1 Bubble sort 232

 5.6.2 Odd-even transposition sort 233

5.7 Scheduling Loops 236

 5.7.1 The schedule clause 237

 5.7.3 The dynamic and guided schedule types 239

 5.7.4 The runtime schedule type 239

 5.7.5 Which schedule? 241

5.8 Producers and Consumers 241

 5.8.1 Queues241

 5.8.2 Message-passing 242

 5.8.3 Sending messages 243

 5.8.4 Receiving messages 243

 5.8.5 Termination detection 244

 5.8.6 Startup 244

 5.8.7 The atomic directive 245

 5.8.8 Critical sections and locks 246

 5.8.9 Using locks in the message-passing program 248

 5.8.10 critical directives, atomic directives, or locks? 249

 5.8.11 Some caveats 249

5.9 Caches, Cache Coherence, and False Sharing 251

5.10 Thread-Safety 256

 5.10.1 Incorrect programs can produce correct output 258

5.11 Summary 259

5.12 Exercises 263

5.13 Programming Assignments .267

CHAPTER 6 Parallel Program Development 271

6.1 Two n-Body Solvers 271

 6.1.1 The problem 271

 6.1.2 Two serial programs 273

 6.1.3 Parallelizing the n-body solvers 277

 6.1.4 A word about I/O 280

 6.1.5 Parallelizing the basic solver using OpenMP 281

 6.1.6 Parallelizing the reduced solver using OpenMP 284

 6.1.7 Evaluating the OpenMP codes 288

 6.1.8 Parallelizing the solvers using pthreads 289

 6.1.9 Parallelizing the basic solver using MPI 290

 6.1.10 Parallelizing the reduced solver using MPI 292

 6.1.11 Performance of the MPI solvers 297

6.2 Tree Search 299

 6.2.1 Recursive depth-first search 302

 6.2.2 Nonrecursive depth-first search 303

 6.2.3 Data structures for the serial implementations 305

 6.2.6 A static parallelization of tree search using pthreads 309

 6.2.7 A dynamic parallelization of tree search using pthreads 310

 6.2.8 Evaluating the pthreads tree-search programs 315

 6.2.9 Parallelizing the tree-search programs using OpenMP 316

 6.2.10 Performance of the OpenMP implementations 318

 6.2.11 Implementation of tree search using MPI and static partitioning 319

 6.2.12 Implementation of tree search using MPI and dynamic partitioning 327

6.3 A Word of Caution 335

6.4 Which API? 335

6.5 Summary 336

 6.5.1 Pthreads and OpenMP 337

 6.5.2 MPI 338

6.6 Exercises 341

6.7 Programming Assignments 350

CHAPTER 7 Where to Go from Here 353

References 357

Index 361