---

"This book is required reading for anyone working with accelerator-based computing systems." From the Foreword by Jack Dongarra, University of Tennessee and Oak Ridge National Laboratory Using NVIDIArs"s breakthrough CUDA software platform, you can harness the immense power of NVIDIA graphics processors to build high-performance, non-graphics software in fields ranging from science and engineering to finance. InCUDA by Example,two senior members of NVIDIArs"s CUDA development team show C/C++ programmers exactly how to make the most of this extraordinary new technology, even if they have absolutely no graphics or parallel programming experience. The only CUDA book created with NVIDIArs"s direct involvement,CUDA by Exampleintroduces every area of CUDA development through working, compilable examples. After concisely introducing the platform and architecture, the authors present a quick-start guide to CUDA C, the C-based language for programming massively parallel NVIDIA GPUs. Next, they systematically detail the techniques and tradeoffs associated with each key CUDA feature. Yours"ll discover when to use each CUDA C extensionand how to write CUDA software that delivers truly outstanding performance. Major topics covered include Parallel programming Thread cooperation Constant memory and events Texture memory Graphics interoperability Atomics Streams CUDA C on Multiple GPUs Advanced atomics

Jason Sanders is a senior software engineer in the CUDA Platform group at NVIDIA. While at NVIDIA, he helped develop early releases of CUDA system software and contributed to the OpenCL 1.0 Specification, an industry standard for heterogeneous computing. Jason received his master’s degree in computer science from the University of California Berkeley where he published research in GPU computing, and he holds a bachelor’s degree in electrical engineering from Princeton University. Prior to joining NVIDIA, he previously held positions at ATI Technologies, Apple, and Novell. When he’s not writing books, Jason is typically working out, playing soccer, or shooting photos.

 

Edward Kandrot is a senior software engineer on the CUDA Algorithms team at NVIDIA. He has more than twenty years of industry experience focused on optimizing code and improving performance, including for Photoshop and Mozilla. Kandrot has worked for Adobe, Microsoft, and Google, and he has been a consultant at many companies, including Apple and Autodesk. When not coding, he can be found playing World of Warcraft or visiting Las Vegas for the amazing food.

Foreword xiii

Preface xv

Acknowledgments xvii

About the Authors xix

 

Chapter 1: Why CUDA? Why Now? 1

1.1 Chapter Objectives 2

1.2 The Age of Parallel Processing 2

1.3 The Rise of GPU Computing 4

1.4 CUDA 6

1.5 Applications of CUDA 8

1.6 Chapter Review 11

 

Chapter 2: Getting Started 13

2.1 Chapter Objectives 14

2.2 Development Environment 14

2.3 Chapter Review 19

 

Chapter 3: Introduction to CUDA C 21

3.1 Chapter Objectives 22

3.2 A First Program 22

3.3 Querying Devices 27

3.4 Using Device Properties 33

3.5 Chapter Review 35

 

Chapter 4: Parallel Programming in CUDA C 37

4.1 Chapter Objectives 38

4.2 CUDA Parallel Programming 38

4.3 Chapter Review 57

 

Chapter 5: Thread Cooperation 59

5.1 Chapter Objectives 60

5.2 Splitting Parallel Blocks 60

5.3 Shared Memory and Synchronization 75

5.4 Chapter Review 94

 

Chapter 6: Constant Memory and Events 95

6.1 Chapter Objectives 96

6.2 Constant Memory 96

6.3 Measuring Performance with Events 108

6.4 Chapter Review 114

 

Chapter 7: Texture Memory 115

7.1 Chapter Objectives 116

7.2 Texture Memory Overview 116

7.3 Simulating Heat Transfer 117

7.4 Chapter Review 137

 

Chapter 8: Graphics Interoperability 139

8.1 Chapter Objectives 140

8.2 Graphics Interoperation 140

8.3 GPU Ripple with Graphics Interoperability 147

8.4 Heat Transfer with Graphics Interop 154

8.5 DirectX Interoperability 160

8.6 Chapter Review 161

 

Chapter 9: Atomics 163

9.1 Chapter Objectives 164

9.2 Compute Capability 164

9.3 Atomic Operations Overview 168

9.4 Computing Histograms 170

9.5 Chapter Review 183

 

Chapter 10: Streams 185

10.1 Chapter Objectives 186

10.2 Page-Locked Host Memory 186

10.3 CUDA Streams 192

10.4 Using a Single CUDA Stream 192

10.5 Using Multiple CUDA Streams 198

10.6 GPU Work Scheduling 205

10.7 Using Multiple CUDA Streams Effectively 208

10.8 Chapter Review 211

 

Chapter 11: CUDA C on Multiple GPUs 213

11.1 Chapter Objectives 214

11.2 Zero-Copy Host Memory 214

11.3 Using Multiple GPUs 224

11.4 Portable Pinned Memory 230

11.5 Chapter Review 235

 

Chapter 12: The Final Countdown 237

12.1 Chapter Objectives 238

12.2 CUDA Tools 238

12.3 Written Resources 244

12.4 Code Resources 246

12.5 Chapter Review 248

 

Appendix A: Advanced Atomics 249

A.1 Dot Product Revisited 250

A.2 Implementing a Hash Table 258

A.3 Appendix Review 277

 

Index 279