Editorial Reviews

Product Description

Foundations of Multithreaded, Parallel, and Distributed Programming covers-and then applies-the core concepts and techniques needed for an introductory course in this topic. The book emphasizes the practice and application of parallel systems, using real-world examples throughout.

Greg Andrews teaches the fundamental concepts of multithreaded, parallel and distributed computing and relates them to the implementation and performance processes. He presents the appropriate breadth of topics and supports these discussions with an emphasis on performance.

From the Back Cover

Greg Andrews teaches the fundamental concepts of multithreaded, parallel and distributed computing and relates them to the implementation and performance processes. He presents the appropriate breadth of topics and supports these discussions with an emphasis on performance.

Features

* Emphasizes how to solve problems, with correctness the primary concern and performance an important, but secondary, concern

* Includes a number of case studies which cover such topics as pthreads, MPI, and OpenMP libraries, as well as programming languages like Java, Ada, high performance Fortran, Linda, Occam, and SR

* Provides examples using Java syntax and discusses how Java deals with monitors, sockets, and remote method invocation

* Covers current programming techniques such as semaphores, locks, barriers, monitors, message passing, and remote invocation

* Concrete examples are executed with complete programs, both shared and distributed

* Sample applications include scientific computing and distributed systems

Preface

Chapter 1: The Concurrent Computing Landscape1

1.1 The Essence of Concurrent Pr-ogrammiilg 2

1.2 Hardware Architectures .

1.2.1 Processors and Caches 4

I 2.2 Shared-Me1nor.y Multiprocessors 6

1.2.3 Distributed-Memory Multicomputers and Networks8

1.3 Applications and Progralnlning Styles 10

1.4 Iterative Parallelism: Matrix Multiplication 13

1.5 Recursive Parallelism: Adaptjve Quadrature17

1.6 Producers and Consumers: Unix Pipes 19

1.7 Clients and Servers: File Systems . 27

1.8 Peers: Distributed Matrix Multiplication 23

1.9 Summary of Programming Notation 26

1.9.1 Declarations26

19.2 Sequential Statements 27

1.9.3 Concurrent Statements, Processes, and Procedures 29

1.9.4 Colnments 31

Historical Notes 31

References 33

.Exercises 34

Part 1 : Shared-Variable Programming

Chapter 2: Processes and Synchronization 41

2.1 States. Actions. Histories. and Properlies42

2.2 Parallelization: Finding Patterns in a File . 44

2.3 Synchronization: The Maximum of an Array 48

. 2.4 Atomic Actions and Await Statements 51

2.4.1 Fine-Grained Atomicity 51

2.4.2 Specifying Synchronization: The Await Statement 54

2.5 Produce~/Consurner Synchronization 56

2.6 A Synopsis of Axiomatic Semantics 57

2.6. 1 Fol.mai Logical Systems 58

2.6.2 A Programming Logic 59

2.6.3 Semantics of Concurrent Execution 62

2.7 Techniques fool- Avoiding Interference 65

2.7.1 Disjoint Variables 65

2.7.2 Weakened Assertions 66

2.7.3 Global Invariants 68

. 2.7.4 S y~ichronizatjon 69

. 2.7.5 An Example: The An-ay Copy f roblern Revisited 70

2.8 Safety and Liveness Properties 72

2.8.1 Proving Safety Properties 73

2.8.2 ScheduJiog Policies and Fairness 74

Historical Notes77

References80

Exercises 81

Chapter 3: Locks and Barriers 93

3.1 The Critical Secrion Problem 94

3.2 Critical Sections: Spin Locks 97

3.2.1 Test and Set 98

3.2.2 Test and Test and Set 100

. 3.2.3 Implementing Await Statements 101

3.3 Critical Sections: Fair Solutions 104

3.3.1 Tfie Tie-Breaker Algorithm 104

3.3.2 The Ticket Algorithm 108

3.3.3 The Bakery Algorithm 11 I.

3.4 Barrier Synchronizatjon 115

3.4.1 Shqred Counter 116

3.4.2 Flags and Coordinators 117

3.4.3 Symmetric Barriers120

3.5 Data Paallel Algorithms 124

3.5.1 Parallel Prefix Computations 124

3.5.2 Operations on Linked Lists127

3.5.3 Grid Compntations: Jacobi Iteration129

3.5.4 Synchronous Multiprocessors 131

3.6 Paral.l.el Computing with a Bag of Tasks 132

3.6.1 Matrix Multiplication133

3.6.2 Adaptive Quadrature 134

. Historical Notes 135

References139

Exercises 141

Chapter 4: Semaphores 153

4 . I Syntax and Semantics 154

4.2 Basic Problems and Techniques156

4.2.1 Critical Sections: Mutual Exclusion156

4.2.2 B tiers: Signaling Events15G

4.2.3 Producers and Consumers: Split Binary Semapl~ores 158

4.2.4 Bounded Buffers: Resource Counting160

4.3 The Dining Philosophers 164

4.4 Readers and Writers 166

4.4.1 ReaderslWriters as an Exclusion Problem 167

4.4.2 Readerstwriters Using Condition Synchronization 169

4.4.3 The Technique of Passing the Baton 171

4.4.4 Alternative Scheduling Policies 175

4.5 Resource Allocation and Scheduling178

. 4.5.1 Problem Definition and General Solution Pattern 178

4.5.2 Shortest-Job-Next Allocation . 180

4.6 Case Study: Pthreads 184

4.6.1 Thread CI-eation 185

4.6.2 Semaphores186

4.6.3 Example: A Simple Producer and Consumer 186

his to^. ical Notes188

References190

E~ercises 191

Chapter 5: Monitors 203

5.1 Syntax and Semanlics 204

5 . 11 Mutual Exclusion 206

5 . 1.2 Condition Variables 207

. 5.1.3 Signaling Disciplines 208

5.1.4 Additional Operations on Condition Variables 212

5.2 Synchronization Techniques 213

. 5.2.1 Bounded Buffers: Basic Condition Synchronization 213

5.2.2 Readers and Writers: Broadcast Signal 215

5.2.3 Shortest-Job-Next Allocation: Priority Wait 21 7

5.2.4 Interval Tii11er: Covering Conditiolls 218

. 5.2.5 The Sleeping Barber: Rendezvous 221

5.3 Disk Scheduling: Program Structures 224

5.3.1 Using a Separate Monitor 228

5.3.2 Using an Intermediary 230

. 5.3.3 Using a Nested Monitor 235

5.4 Case Study: Java 237

5.4.1 The Tl~reads Class 238

5.4.2 Synchonized Methods 239

5.4.3 Parallel ReadersIWriters 241

. 5.4.4 Exclusive ReadersNriters 243

. 5.4.5 True ReadersIWriters 245

5.5 Case Study: f theads 246

5.5.1 Locks and Condition Variables 246

5.5.2 Example: Summing the Elements of a Matrix 248

Historical Notes 250

References 253

. Exercises 255

Chapter 6: Implementations 265

. 6 .I A Single-Pi-ocessor Kernel 7-66

6.2 A Multiprocessor. Kernel 270

6.3 Implementing Semaphores in a Kernel 276

6.4 Impleinenting Monitors in a Kernel 279

6.5 Implementing Monitors Using Semaphores 283

HistoricaI Notes 284

'XF, References -.

Exercises 137

Part 2: Distributed Programming

. Chapter 7: Message Passing 29s

7.1 Asynchronous Message Passing 296

7.2 Filters: A Sor-ring Network298

7.3 Clients and Servers 302

7.3.1 Active IVlonitors302

7.3.2 A Self-scheduling Disk Server 308

7.3.3 File Servers: Conversational Continuity 311

7.4 Interacting Peers: Exchanging Values 314

7.5 Synchronous Message Passing 318

7.6 Case Study: CSP 320

7.6.1 Corn~nunication Statements 321

7.6.2 Guarded Communication 323

7.6.3 Example: The Sieve of Eratosthenes 326

7.6.4 Occam and Modern CSP 328

7.7 Case Study: Linda 334

7.7.1 Tuple Space and Process Interaction 334

7.7.2 Example: Prime Numbers with a Bag of Tasks337

7.8 Case Study: MPI 340

7.8.1 Basic Functions 341

7.8.2 Global Communication and Synchronization 343

7.9 Case Study: Java 344

7.9.1 Networks and Sockets 344

. 7.9.2 Example: A Remote File Reader 345

Histol-icai Notes 348

References 351

Exercises353

Chapter 8: RPC and Rendezvous 361

8.1 Remote Procedure Call362

8.11 Synchronization in Modules 364

8.1.2 A Time Server365

8.1.3 Caches in a Distributed File System 367

8.1.4 A Sorting Network of Merge Filters 370

8.1.5 Interacting Peers: Exchanging Values 371

8.2 Rendezvous 373

8.2.1 Input Statements374

8.2.2 ClienZ/Server Examples376

8.2.3 A Sorting Network of Merge Filters 379

8.2.4 Interacting Peers: Exchanging Values 381

8.3 A Multiple Primitives Notation 382

8.3.1 lnvoking and Servicing Operations 382

8.3.2 Examples384

5.4 Readers~Writers Revisited 386

8.4.1 Encapsulated Access 387

8.4.2 Replicated Files 389

8.5 Case Study: Java 393

8.5. 1 Remote Method Invocation 393

8.5.2 Example: A Remote Database 39.5

8.6 Case Study: Ada 397

. 8.6.1 Tasks 398

. 8.6.2 Rendezvous 399

5.6.3 ProteccedTypes 401

8.6.4 Example: The Dining Philosophers 403

8.7 Case Study: SR 406

8.7.1 Resources and Globals 406

8.7.2 Comrnunication and Synchronjzation 405

8.7.3 Example: Critical Section Silnulation 409

Historical Notes 411

References 41.5

Exercises 416

Chapter 9: Paradigms for Process Interaction 423

9.1 Manager/Worlzers (Distributed Bag of Tasks) 424

. 9 . I1 Spai-se Matrix Multiplication 424

9.1.2 Adaptive Quadrature Revisited 428

9.2 Heartbeat Algorithms . 430

9.2.1. Image Processing: Region Labeling 432

9.2.2 Cellular Automata: The Game of Life 435

9.3 Pipeline Algorithms 437

9.3.1 A Distribu~ed Matrix Multiplication Pipeliile 438

0.3.2 Matrix Multiplication by Blocks 441

9.4 ProbeEcho Algorithms 444

9.4.1 Broadcast in a Network 444

9.4.2 Computing the Topology of a Network448

9.5 Broadcast Algorithms 451

9.5.1 Logical Clocks and Event Ordering 452

. 9.5.2 Distributed Semaphores 454

. 9.6 Token-Passing Algoritl~~ns 457

9.6.1 Distributed Mutual Excl~~sion 457

9.6.2 Termination Derectio~z in a Ring 460

9.6.3 Termination Detection in a Graph 462

9.7 Replicated Servers465

9.7.1 Distributed Dining Philosophers 466

9.7.2 Decentralized Dining Philosophers 407

Historical Notes 471

References474

Exercises 477

Chapter 10: Implementations 487

. 10.1 Asynchronous Message Passing 488

10.1. I Shared-Memory Keniel 488

10.1.2 Distributed Kernel 491

10.2 Synchronous Message Passing 496

10.2.1. Direct Communication Using Asynchronous Messages 497

10.2.2 Guarded Com~nunication Using a Clearinghouse 498

10.3 RPC and Rendezvous504

10.3.1 RPC in a Kernel 504

10.3.2 Rendezvous Using Asynchronous Message Passing507

10.3.3 Multiple Primitives in a Kernel 509

10.4 Distributed Shared Memory515

10.4.1 Implementation Overview516

10.4.2 Page Consistency Protocols 518

Historical Notes 520

References521

Exercises 522

. Part 3: Parallel Programming 527

Chapter 11 : Scientific Computing533

. 11.1 Grid Cornputations 534

11.1.1 Laplace's Equation 534

11.1.2 Sequential Jacobi Iteration535

11 .I 3 Jacobi Iteration Using Shared Variables 540

I 1.1.4 Jacobi Iteration Using Message Passing 541

1 1.1.5 Red/Black Successive Over-Relaxation (SOR)546

1 1.1.6 Multigrid Methods549

11.2 Particle Computations 553

I 1.2.1 The Gravitatioilal N-Body Problem 554

1 1.2.2 Shared-Variable Program 555

1 1.2.3 Message-Passing Programs 559

1 1.2.4 Approximate Methods 569

11.3 Matrix Computations 573

11.3.1 Gaussian Elimination 573

11.3.2 LU Decomposition575

1 1.3.3 Shared-Variable Program 576

1 1.3.4 Message-Passing Program 581

Historical Notes 583

References 584

Exercises 585

Chapter 12: Languages. Compilers.

Libraries. and Tools 591

12.1 Parallel Programming Libraries592

12.1.1 Case Study: Pthreads 593

12.1.2 Case Study: MPI 593

. 12.1.3 Case Study: OpenMP 595

12.2 Parallelizing Compilers 603

12.2.1 Dependence Analysis604

12.2.2 Prograrn Transfo~mations 607

12.3 Languages and Models 614

12.3. 1 Imperative Languages . 616

12.3.2 Coordination Languages 619

12.3.3 Data Parallel Languages 620

12.3.4 Functional Languages 623

12.3.5 Abstracr Models626

12.3.6 Case Study: High-Performance Fortran (HPF) 629

12.4 Parallel Prograinming Tools 633

12.4.1 Pel-formance Measurement and Visualization 633

12.4.2 Metacornpilters and Metacornpuling 634

12.4.3 Case Study: The Globus Toolkit 636

Historical Notes 638

References642

Exercises 644

Glossary647

l ndex 657