GPU Gems 3

Foreword

Composition, the organization of elemental operations into a nonobvious whole, is the essence of imperative programming. The instruction set architecture (ISA) of a microprocessor is a versatile composition interface, which programmers of software renderers have used effectively and creatively in their quest for image realism. Early graphics hardware increased rendering performance, but often at a high cost in composability, and thus in programmability and application innovation. Hardware with microprocessor-like programmability did evolve (for example, the Ikonas Graphics System), but the dominant form of graphics hardware acceleration has been organized around a fixed sequence of rendering operations, often referred to as the graphics pipeline. Early interfaces to these systems—such as CORE and later, PHIGS—allowed programmers to specify rendering results, but they were not designed for composition.

OpenGL, which I helped to evolve from its Silicon Graphics-defined predecessor IRIS GL in the early 1990s, addressed the need for composability by specifying an architecture (informally called the OpenGL Machine) that was accessed through an imperative programmatic interface. Many features—for example, tightly specified semantics; table-driven operations such as stencil and depth-buffer functions; texture mapping exposed as a general 1D, 2D, and 3D lookup function; and required repeatability properties—ensured that programmers could compose OpenGL operations with powerful and reliable results. Some of the useful techniques that OpenGL enabled include texture-based volume rendering, shadow volumes using stencil buffers, and constructive solid geometry algorithms such as capping (the computation of surface planes at the intersections of clipping planes and solid objects defined by polygons). Ultimately, Mark Peercy and the coauthors of the SIGGRAPH 2000 paper "Interactive Multi-Pass Programmable Shading" demonstrated that arbitrary RenderMan shaders could be accelerated through the composition of OpenGL rendering operations.

During this decade, increases in the raw capability of integrated circuit technology allowed the OpenGL architecture (and later, Direct3D) to be extended to expose an ISA interface. These extensions appeared as programmable vertex and fragment shaders within the graphics pipeline and now, with the introduction of CUDA, as a data-parallel ISA in near parity with that of the microprocessor. Although the cycle toward complete microprocessor-like versatility is not complete, the tremendous power of graphics hardware acceleration is more accessible than ever to programmers.

And what computational power it is! At this writing, the NVIDIA GeForce 8800 Ultra performs over 400 billion floating-point operations per second—more than the most powerful supercomputer available a decade ago, and five times more than today's most powerful microprocessor. The data-parallel programming model the Ultra supports allows its computational power to be harnessed without concern for the number of processors employed. This is critical, because while today's Ultra already includes over 100 processors, tomorrow's will include thousands, and then more. With no end in sight to the annual compounding of integrated circuit density known as Moore's Law, massively parallel systems are clearly the future of computing, with graphics hardware leading the way.

GPU Gems 3 is a collection of state-of-the-art GPU programming examples. It is about putting data-parallel processing to work. The first four sections focus on graphics-specific applications of GPUs in the areas of geometry, lighting and shadows, rendering, and image effects. Topics in the fifth and sixth sections broaden the scope by providing concrete examples of nongraphical applications that can now be addressed with data-parallel GPU technology. These applications are diverse, ranging from rigid-body simulation to fluid flow simulation, from virus signature matching to encryption and decryption, and from random number generation to computation of the Gaussian.

Where is this all leading? The cover art reminds us that the mind remains the most capable parallel computing system of all. A long-term goal of computer science is to achieve and, ultimately, to surpass the capabilities of the human mind. It's exciting to think that the computer graphics community, as we identify, address, and master the challenges of massively parallel computing, is contributing to the realization of this dream.

Kurt Akeley
Microsoft Research

• Contributors
• Foreword
• Part I: Geometry
  • Chapter 1. Generating Complex Procedural Terrains Using the GPU
  • Chapter 2. Animated Crowd Rendering
  • Chapter 3. DirectX 10 Blend Shapes: Breaking the Limits
  • Chapter 4. Next-Generation SpeedTree Rendering
  • Chapter 5. Generic Adaptive Mesh Refinement
  • Chapter 6. GPU-Generated Procedural Wind Animations for Trees
  • Chapter 7. Point-Based Visualization of Metaballs on a GPU
• Part II: Light and Shadows
  • Chapter 10. Parallel-Split Shadow Maps on Programmable GPUs
  • Chapter 11. Efficient and Robust Shadow Volumes Using Hierarchical Occlusion Culling and Geometry Shaders
  • Chapter 12. High-Quality Ambient Occlusion
  • Chapter 13. Volumetric Light Scattering as a Post-Process
  • Chapter 8. Summed-Area Variance Shadow Maps
  • Chapter 9. Interactive Cinematic Relighting with Global Illumination
• Part III: Rendering
  • Chapter 14. Advanced Techniques for Realistic Real-Time Skin Rendering
  • Chapter 15. Playable Universal Capture
  • Chapter 16. Vegetation Procedural Animation and Shading in Crysis
  • Chapter 17. Robust Multiple Specular Reflections and Refractions
  • Chapter 18. Relaxed Cone Stepping for Relief Mapping
  • Chapter 19. Deferred Shading in Tabula Rasa
  • Chapter 20. GPU-Based Importance Sampling
• Part IV: Image Effects
  • Chapter 21. True Impostors
  • Chapter 22. Baking Normal Maps on the GPU
  • Chapter 23. High-Speed, Off-Screen Particles
  • Chapter 24. The Importance of Being Linear
  • Chapter 25. Rendering Vector Art on the GPU
  • Chapter 26. Object Detection by Color: Using the GPU for Real-Time Video Image Processing
  • Chapter 27. Motion Blur as a Post-Processing Effect
  • Chapter 28. Practical Post-Process Depth of Field
• Part V: Physics Simulation
  • Chapter 29. Real-Time Rigid Body Simulation on GPUs
  • Chapter 30. Real-Time Simulation and Rendering of 3D Fluids
  • Chapter 31. Fast N-Body Simulation with CUDA
  • Chapter 32. Broad-Phase Collision Detection with CUDA
  • Chapter 33. LCP Algorithms for Collision Detection Using CUDA
  • Chapter 34. Signed Distance Fields Using Single-Pass GPU Scan Conversion of Tetrahedra
  • Chapter 35. Fast Virus Signature Matching on the GPU
• Part VI: GPU Computing
  • Chapter 36. AES Encryption and Decryption on the GPU
  • Chapter 37. Efficient Random Number Generation and Application Using CUDA
  • Chapter 38. Imaging Earth's Subsurface Using CUDA
  • Chapter 39. Parallel Prefix Sum (Scan) with CUDA
  • Chapter 40. Incremental Computation of the Gaussian
  • Chapter 41. Using the Geometry Shader for Compact and Variable-Length GPU Feedback
• Preface