GPU Gems 3 Book

Foreword xxviiPreface xxixContributors xxxiiiPART I: GEOMETRY 3Chapter 1: Generating Complex Procedural Terrains Using the GPU 7
Ryan Geiss, NVIDIA Corporation

1.1 Introduction 7

1.2 Marching Cubes and the Density Function 7

1.3 An Overview of the Terrain Generation System 12

1.4 Generating the Polygons Within a Block of Terrain 20

1.5 Texturing and Shading 29

1.6 Considerations for Real-World Applications 35

1.7 Conclusion 37

1.8 References 37Chapter 2: Animated Crowd Rendering 39
Bryan Dudash, NVIDIA Corporation

2.1 Motivation 39

2.2 A Brief Review of Instancing 40

2.3 Details of the Technique 42

2.4 Other Considerations 50

2.5 Conclusion 51

2.6 References 52Chapter 3: DirectX 10 Blend Shapes: Breaking the Limits 53
Tristan Lorach, NVIDIA Corporation

3.1 Introduction 53

3.2 How Does It Work? 56

3.3 Running the Sample 66

3.4 Performance 66

3.5 References 67Chapter 4: Next-Generation SpeedTree Rendering 69
Alexander Kharlamov, NVIDIA Corporation
Iain Cantlay, NVIDIA Corporation
Yury Stepanenko, NVIDIA Corporation

4.1 Introduction 69

4.2 Silhouette Clipping 69

4.3 Shadows 76

4.4 Leaf Lighting 81

4.5 High Dynamic Range and Antialiasing 85

4.6 Alpha to Coverage 85

4.7 Conclusion 88

4.8 References 91Chapter 5: Generic Adaptive Mesh Refinement 93
Tamy Boubekeur, LaBRI–INRIA, University of Bordeaux
Christophe Schlick, LaBRI–INRIA, University of Bordeaux

5.1 Introduction 94

5.2 Overview 95

5.3 Adaptive Refinement Patterns 96

5.4 Rendering Workflow 98

5.5 Results 100

5.6 Conclusion and Improvements 103

5.7 References 104Chapter 6: GPU-Generated ProceduralWind Animations for Trees 105
Renaldas Zioma, Electronic Arts/Digital Illusions CE

6.1 Introduction 105

6.2 Procedural Animations on the GPU 106

6.3 A Phenomenological Approach 106

6.4 The Simulation Step 113

6.5 Rendering the Tree 117

6.6 Analysis and Comparison 118

6.7 Summary 119

6.8 References 120Chapter 7: Point-Based Visualization of Metaballs on a GPU 123
Kees van Kooten, Playlogic Game Factory
Gino van den Bergen, Playlogic Game Factory
Alex Telea, Eindhoven University of Technology

7.1 Metaballs, Smoothed Particle Hydrodynamics, and Surface Particles 124

7.2 Constraining Particles 127

7.3 Local Particle Repulsion 135

7.4 Global Particle Dispersion 140

7.5 Performance 145

7.6 Rendering 146

7.7 Conclusion 147

7.8 References 148PART II: LIGHT AND SHADOWS 151Chapter 8: Summed-Area Variance Shadow Maps 157
Andrew Lauritzen, University of Waterloo

8.1 Introduction 157

8.2 Related Work 158

8.3 Percentage-Closer Filtering 159

8.4 Variance Shadow Maps 161

8.5 Summed-Area Variance Shadow Maps 174

8.6 Percentage-Closer Soft Shadows 178

8.7 Conclusion 181

8.8 References 181Chapter 9: Interactive Cinematic Relighting with Global Illumination 183
Fabio Pellacini, Dartmouth College
Miloš Hašan, Cornell University
Kavita Bala, Cornell University

9.1 Introduction 183

9.2 An Overview of the Algorithm 184

9.3 Gather Samples 186

9.4 One-Bounce Indirect Illumination 188

9.5 Wavelets for Compression 189

9.6 Adding Multiple Bounces 192

9.7 Packing Sparse Matrix Data 193

9.8 A GPU-Based Relighting Engine 195

9.9 Results 200

9.10 Conclusion 201

9.11 References 201Chapter 10: Parallel-Split Shadow Maps on Programmable GPUs 203
Fan Zhang, The Chinese University of Hong Kong
Hanqiu Sun, The Chinese University of Hong Kong
Oskari Nyman, Helsinki University of Technology

10.1 Introduction 203

10.2 The Algorithm 205

10.3 Hardware-Specific Implementations 214

10.4 Further Optimizations 232

10.5 Results 233

10.6 Conclusion 233

10.7 References 235Chapter 11: Efficient and Robust Shadow Volumes Using Hierarchical Occlusion Culling and Geometry Shaders 239
Martin Stich, mental images
Carsten Wächter, Ulm University
Alexander Keller, Ulm University

11.1 Introduction 239

11.2 An Overview of Shadow Volumes 240

11.3 Our Implementation 244

11.4 Conclusion 254

11.5 References 254Chapter 12: High-Quality Ambient Occlusion 257
Jared Hoberock, University of Illinois at Urbana-Champaign
Yuntao Jia, University of Illinois at Urbana-Champaign

12.1 Review 257

12.2 Problems 258

12.3 A Robust Solution 261

12.4 Results 267

12.5 Performance 269

12.6 Caveats 270

12.7 Future Work 273

12.8 References 274Chapter 13: Volumetric Light Scattering as a Post-Process 275
Kenny Mitchell, Electronic Arts

13.1 Introduction 275

13.2 Crepuscular Rays 276

13.3 Volumetric Light Scattering 277

13.4 The Post-Process Pixel Shader 279

13.5 Screen-Space Occlusion Methods 281

13.6 Caveats 282

13.7 The Demo 283

13.8 Extensions 284

13.9 Summary 284

13.10 References 284PART III: RENDERING 287Chapter 14: Advanced Techniques for Realistic Real-Time Skin Rendering 293
Eugene d’Eon, NVIDIA Corporation
David Luebke, NVIDIA Corporation

14.1 The Appearance of Skin 293

14.2 An Overview of the Skin-Rendering System 297

14.3 Specular Surface Reflectance 299

14.4 Scattering Theory 305

14.5 Advanced Subsurface Scattering 314

14.6 A Fast Bloom Filter 342

14.7 Conclusion 342

14.8 References 345Chapter 15: Playable Universal Capture 349
George Borshukov, Electronic Arts
Jefferson Montgomery, Electronic Arts
John Hable, Electronic Arts

15.1 Introduction 349

15.2 The Data Acquisition Pipeline 350

15.3 Compression and Decompression of the Animated Textures 352

15.4 Sequencing Performances 363

15.5 Conclusion 363

15.6 References 370Chapter 16: Vegetation Procedural Animation and Shading in Crysis 373
Tiago Sousa, Crytek

16.1 Procedural Animation 373

16.2 Vegetation Shading 378

16.3 Conclusion 384

16.4 References 384Chapter 17: Robust Multiple Specular Reflections and Refractions 387
Tamás Umenhoffer, Budapest University of Technology and Economics
Gustavo Patow, University of Girona
László Szirmay-Kalos, Budapest University of Technology and Economics

17.1 Introduction 388

17.2 Tracing Secondary Rays 389

17.3 Reflections and Refractions 396

17.4 Results 400

17.5 Conclusion 402

17.6 References 406Chapter 18: Relaxed Cone Stepping for Relief Mapping 409
Fabio Policarpo, Perpetual Entertainment
Manuel M. Oliveira, Instituto de Informática—UFRGS

18.1 Introduction 409

18.2 A Brief Review of Relief Mapping 411

18.3 Cone Step Mapping 415

18.4 Relaxed Cone Stepping 416

18.5 Conclusion 425

18.6 References 427Chapter 19: Deferred Shading in Tabula Rasa 429
Rusty Koonce, NCsoft Corporation

19.1 Introduction 429

19.2 Some Background 430

19.3 Forward Shading Support 431

19.4 Advanced Lighting Features 434

19.5 Benefits of a Readable Depth and Normal Buffer 440

19.6 Caveats 445

19.7 Optimizations 448

19.8 Issues 450

19.9 Results 454

19.10 Conclusion 454

19.11 References 457Chapter 20: GPU-Based Importance Sampling 459
Mark Colbert, University of Central Florida
Jaroslav Křivánek, Czech Technical University in Prague

20.1 Introduction 459

20.2 Rendering Formulation 459

20.3 Quasirandom Low-Discrepancy Sequences 465

20.4 Mipmap Filtered Samples 466

20.5 Performance 470

20.6 Conclusion 471

20.7 Further Reading and References 474PART IV: IMAGE EFFECTS 477Chapter 21: True Impostors 481
Eric Risser, University of Central Florida

21.1 Introduction 481

21.2 Algorithm and Implementation Details 482

21.3 Results 487

21.4 Conclusion 48921.5 References 489Chapter 22: Baking Normal Maps on the GPU 491
Diogo Teixeira, Move Interactive

22.1 The Traditional Implementation 492

22.2 Acceleration Structures 493

22.3 Feeding the GPU 496

22.4 Implementation 498

22.5 Results 508

22.6 Conclusion 511

22.7 References 511Chapter 23: High-Speed, Off-Screen Particles 513
Iain Cantlay, NVIDIA Corporation

23.1 Motivation 513

23.2 Off-Screen Rendering 514

23.3 Downsampling Depth 517

23.4 Depth Testing and Soft Particles 519

23.5 Alpha Blending 520

23.6 Mixed-Resolution Rendering 522

23.7 Results 525

23.8 Conclusion 527

23.9 References 528Chapter 24: The Importance of Being Linear 529
Larry Gritz, NVIDIA Corporation
Eugene d’Eon, NVIDIA Corporation

24.1 Introduction 529

24.2 Light, Displays, and Color Spaces 529

24.3 The Symptoms 533

24.4 The Cure 538

24.5 Conclusion 541

24.6 Further Reading 542Chapter 25: Rendering Vector Art on the GPU 543
Charles Loop, Microsoft Research
Jim Blinn, Microsoft Research

25.1 Introduction 543

25.2 Quadratic Splines 544

25.3 Cubic Splines 546

25.4 Triangulation 555

25.5 Antialiasing 556

25.6 Code 558

25.7 Conclusion 559

25.8 References 560Chapter 26: Object Detection by Color: Using the GPU for Real-Time Video Image Processing 563
Ralph Brunner, Apple
Frank Doepke, Apple
Bunny Laden, Apple

26.1 Image Processing Abstracted 564

26.2 Object Detection by Color 567

26.3 Conclusion 574

26.4 Further Reading 574Chapter 27: Motion Blur as a Post-Processing Effect 575
Gilberto Rosado, Rainbow Studios

27.1 Introduction 575

27.2 Extracting Object Positions from the Depth Buffer 576

27.3 Performing the Motion Blur 579

27.4 Handling Dynamic Objects 580

27.5 Masking Off Objects 580

27.6 Additional Work 581

27.7 Conclusion 581

27.8 References 581Chapter 28: Practical Post-Process Depth of Field 583
Earl Hammon, Jr., Infinity Ward

28.1 Introduction 583

28.2 Related Work 583

28.3 Depth of Field 585

28.4 Evolution of the Algorithm 587

28.5 The Complete Algorithm 592

28.6 Conclusion 602

28.7 Limitations and Future Work 603

28.8 References 605PART V: PHYSICS SIMULATION 607Chapter 29: Real-Time Rigid Body Simulation on GPUs 611
Takahiro Harada, University of Tokyo

29.1 Introduction 613

29.2 Rigid Body Simulation on the GPU 618

29.3 Applications 627

29.4 Conclusion 629

29.5 Appendix 631

29.6 References 631Chapter 30: Real-Time Simulation and Rendering of 3D Fluids 633
Keenan Crane, University of Illinois at Urbana-Champaign
Ignacio Llamas, NVIDIA Corporation
Sarah Tariq, NVIDIA Corporation

30.1 Introduction 633

30.2 Simulation 634

30.3 Rendering 665

30.4 Conclusion 672

30.5 References 673Chapter 31: Fast N-Body Simulation with CUDA 677
Lars Nyland, NVIDIA Corporation
Mark Harris, NVIDIA Corporation
Jan Prins, University of North Carolina at Chapel Hill

31.1 Introduction 677

31.2 All-Pairs N-Body Simulation 679

31.3 A CUDA Implementation of the All-Pairs N-Body Algorithm 680

31.4 Performance Results 686

31.5 Previous Methods Using GPUs for N-Body Simulation 691

31.6 Hierarchical N-Body Methods 692

31.7 Conclusion 693

31.8 References 694Chapter 32: Broad-Phase Collision Detection with CUDA 697
Scott Le Grand, NVIDIA Corporation

32.1 Broad-Phase Algorithms 697

32.2 A CUDA Implementation of Spatial Subdivision 702

32.3 Performance Results 719

32.4 Conclusion 721

32.5 References 721Chapter 33: LCP Algorithms for Collision Detection Using CUDA 723
Peter Kipfer, Havok

33.1 Parallel Processing 724

33.2 The Physics Pipeline 724

33.3 Determining Contact Points 726

33.4 Mathematical Optimization 728

33.5 The Convex Distance Calculation 731

33.6 The Parallel LCP Solution Using CUDA 732

33.7 Results 738

33.8 References 739Chapter 34: Signed Distance Fields Using Single-Pass GPU Scan Conversion of Tetrahedra 741
Kenny Erleben, University of Copenhagen
Henrik Dohlmann, 3Dfacto R&D

34.1 Introduction 741

34.2 Leaking Artifacts in Scan Methods 742

34.3 Our Tetrahedra GPU Scan Method 747

34.4 Results 756

34.5 Conclusion 758

34.6 Future Work 759

34.7 Further Reading 760

34.8 References 762PART VI: GPU COMPUTING 765Chapter 35: Fast Virus Signature Matching on the GPU 771
Elizabeth Seamans, Juniper Networks
Thomas Alexander, Polytime

35.1 Introduction 771

35.2 Pattern Matching 773

35.3 The GPU Implementation 775

35.4 Results 779

35.5 Conclusions and Future Work 782

35.6 References 783Chapter 36: AES Encryption and Decryption on the GPU 785
Takeshi Yamanouchi, SEGA Corporation

36.1 New Functions for Integer Stream Processing 786

36.2 An Overview of the AES Algorithm 788

36.3 The AES Implementation on the GPU 790

36.4 Performance 797

36.5 Considerations for Parallelism 799

36.6 Conclusion and Future Work 802

36.7 References 802Chapter 37: Efficient Random Number Generation and Application Using CUDA 805
Lee Howes, Imperial College London
David Thomas, Imperial College London

37.1 Monte Carlo Simulations 806

37.2 Random Number Generators 809

37.3 Example Applications 821

37.4 Conclusion 829

37.5 References 829Chapter 38: Imaging Earth’s Subsurface Using CUDA 831
Bernard Deschizeaux, CGGVeritas
Jean-Yves Blanc, CGGVeritas

38.1 Introduction 831

38.2 Seismic Data 832

38.3 Seismic Processing 834

38.4 The GPU Implementation 841

38.5 Performance 849

38.6 Conclusion 849

38.7 References 850Chapter 39: Parallel Prefix Sum (Scan) with CUDA 851
Mark Harris, NVIDIA Corporation
Shubhabrata Sengupta, University of California, Davis
John D. Owens, University of California, Davis

39.1 Introduction 851

39.2 Implementation 853

39.3 Applications of Scan 866

39.4 Conclusion 875

39.5 References 875Chapter 40: Incremental Computation of the Gaussian 877
Ken Turkowski, Adobe Systems

40.1 Introduction and Related Work 877

40.2 Polynomial Forward Differencing 879

40.3 The Incremental Gaussian Algorithm 882

40.4 Error Analysis 885

40.5 Performance 887

40.6 Conclusion 888

40.7 References 888Chapter 41: Using the Geometry Shader for Compact and Variable-Length GPU Feedback 891
Franck Diard, NVIDIA Corporation

41.1 Introduction 891

41.2 Why Use the Geometry Shader? 892

41.3 Dynamic Output with the Geometry Shader 893

41.4 Algorithms and Applications 895

41.5 Benefits: GPU Locality and SLI 903

41.6 Performance and Limits 905

41.7 Conclusion 907

41.8 References 907Index 909