book

GPU Gems 3

Name: GPU Gems 3
Author: Hubert Nguyen
ISBN: 9780321545428

by Hubert Nguyen

August 2007

Intermediate to advanced

1008 pages

21h 13m

English

Addison-Wesley Professional

Read now

Unlock full access

Copyright
Foreword
Preface
Intended AudienceTrying the Code SamplesAcknowledgments
Contributors
I. Geometry
1. Generating Complex Procedural Terrains Using the GPU
1.1. Introduction1.2. Marching Cubes and the Density Function1.2.1. Generating Polygons Within a Cell1.2.2. Lookup Tables1.3. An Overview of the Terrain Generation System1.3.1. Generating the Polygons Within a Block of Terrain1.3.2. Generating the Density Values1.3.3. Making an Interesting Density FunctionSampling Tips1.3.4. Customizing the TerrainUse a Hand-Painted 2D TextureAdd Manually Controlled InfluencesAdd Miscellaneous Effects1.4. Generating the Polygons Within a Block of Terrain1.4.1. Margin Data1.4.2. Generating a Block: Method 11.4.3. Generating a Block: Method 21.4.4. Generating a Block: Method 31.5. Texturing and Shading1.6. Considerations for Real-World Applications1.6.1. Level of Detail1.6.2. Collisions and Lighting of Foreign ObjectsCollisionsLighting1.7. Conclusion1.8. References
2. Animated Crowd Rendering
2.1. Motivation2.2. A Brief Review of Instancing2.3. Details of the TechniqueCPUGPU2.3.1. Constants-Based InstancingSV_InstanceID2.3.2. Palette Skinning Using an Animation TextureDecode Matrices from a TextureConditional Branching for WeightsA Few Important Points2.3.3. Geometry Variations2.3.4. The Level-of-Detail System2.4. Other Considerations2.4.1. Color Variations2.4.2. Performance2.4.3. Integration2.5. Conclusion2.6. References
3. DirectX 10 Blend Shapes: Breaking the Limits
3.1. Introduction3.2. How Does It Work?3.2.1. Features of DirectX 103.2.2. Defining the Mesh3.2.3. The Stream-Out Method3.2.4. The Buffer-Template Method3.3. Running the Sample3.4. Performance3.5. References

4. Next-Generation SpeedTree Rendering
4.1. Introduction4.2. Silhouette Clipping4.2.1. Silhouette Fin Extrusion4.2.2. Height Tracing4.2.3. Silhouette Level of Detail4.3. Shadows4.3.1. Leaf Self-Shadowing4.3.2. Cascaded Shadow Mapping4.4. Leaf Lighting4.4.1. Two-Sided Lighting4.4.2. Specular Lighting4.5. High Dynamic Range and Antialiasing4.6. Alpha to Coverage4.6.1. Alpha to Coverage Applied to SpeedTrees4.6.2. Level-of-Detail Cross-Fading4.6.3. Silhouette Edge Antialiasing4.7. Conclusion4.8. References
5. Generic Adaptive Mesh Refinement
5.1. Introduction5.2. Overview5.3. Adaptive Refinement Patterns5.3.1. Implementation5.4. Rendering Workflow5.4.1. Depth-Tagging5.4.2. The CPU-Level Rendering Loop5.4.3. The GPU-Level Refinement Process5.5. Results5.6. Conclusion and Improvements5.7. References
6. GPU-Generated Procedural Wind Animations for Trees
6.1. Introduction6.2. Procedural Animations on the GPU6.3. A Phenomenological Approach6.3.1. The Wind Field6.3.2. The Conceptual Structure of a Tree6.3.3. The Two Categories of SimulationAnimating the TrunkAnimating the BranchesImprovements for Branch Animations6.4. The Simulation Step6.4.1. The Quaternion Library in HLSL6.5. Rendering the Tree6.5.1. DirectX 106.6. Analysis and Comparison6.6.1. Pros6.6.2. Cons6.6.3. Performance Results6.7. Summary6.8. References
7. Point-Based Visualization of Metaballs on a GPU
7.1. Metaballs, Smoothed Particle Hydrodynamics, and Surface Particles7.1.1. A Comparison of Methods7.1.2. Point-Based Surface Visualization on a GPU7.2. Constraining Particles7.2.1. Defining the Implicit Surface7.2.2. The Velocity Constraint Equation7.2.3. Computing the Density Field on the GPU7.2.4. Choosing the Hash Function7.2.5. Constructing and Querying the Hash7.3. Local Particle Repulsion7.3.1. The Repulsion Force Equation7.3.3. Nearest Neighbors on a GPU7.4. Global Particle Dispersion7.5. Performance7.6. Rendering7.7. Conclusion7.8. References
II. Light and Shadows
8. Summed-Area Variance Shadow Maps
8.1. Introduction8.2. Related Work8.3. Percentage-Closer Filtering8.3.1. Problems with Percentage-Closer Filtering8.4. Variance Shadow Maps8.4.1. Filtering the Variance Shadow Map8.4.2. Biasing8.4.3. Light BleedingAn Approximate Algorithm (Light-Bleeding Reduction)8.4.4. Numeric Stability8.4.5. Implementation Notes8.4.6. Variance Shadow Maps and Soft Shadows8.5. Summed-Area Variance Shadow Maps8.5.1. Generating Summed-Area Tables8.5.2. Numeric Stability Revisited8.5.3. Results8.6. Percentage-Closer Soft Shadows8.6.1. The Blocker Search8.6.2. Penumbra Size Estimation8.6.3. Shadow Filtering8.6.4. Results8.7. Conclusion8.8. References
9. Interactive Cinematic Relighting with Global Illumination
9.1. Introduction9.2. An Overview of the Algorithm9.3. Gather Samples9.4. One-Bounce Indirect Illumination9.5. Wavelets for Compression9.6. Adding Multiple Bounces9.7. Packing Sparse Matrix Data9.8. A GPU-Based Relighting Engine9.8.1. Direct Illumination9.8.2. Wavelet Transform9.8.3. Sparse Matrix Multiplication9.9. Results9.10. Conclusion9.11. References
10. Parallel-Split Shadow Maps on Programmable GPUs
10.1. Introduction10.2. The Algorithm10.2.1. Step 1: Splitting the View FrustumShadow-Map AliasingThe Practical Split SchemePreprocessing10.2.2. Step 2: Calculating Light’s Transformation MatricesScene-Independent ProjectionScene-Dependent Projection10.2.3. Steps 3 and 4: Generating PSSMs and Synthesizing Shadows10.3. Hardware-Specific Implementations10.3.1. The Multipass MethodGenerating Shadow MapsSynthesizing Shadows10.3.2. DirectX 9-Level AccelerationThe SetupSynthesizing ShadowsPerformance10.3.3. DirectX 10-Level AccelerationFeatures in Direct3D 10The SetupGenerating Shadow MapsUsing Geometry Shader CloningUsing InstancingSynthesizing ShadowsUsing Cube Maps10.4. Further Optimizations10.5. Results10.6. Conclusion10.7. References
11. Efficient and Robust Shadow Volumes Using Hierarchical Occlusion Culling and Geometry Shaders
11.1. Introduction11.2. An Overview of Shadow Volumes11.2.1. Z-Pass and Z-Fail11.2.2. Volume GenerationRendering StepsRendering at Infinity11.2.3. Performance and Optimizations11.3. Our Implementation11.3.1. Robust Shadows for Low-Quality MeshesA Modified Volume Generation AlgorithmPerformance Costs11.3.2. Dynamic Volume Generation with Geometry Shaders11.3.3. Improving Performance with Hierarchical Occlusion Culling11.4. Conclusion11.5. References
12. High-Quality Ambient Occlusion
12.1. Review12.2. Problems12.2.1. Disk-Shaped Artifacts12.2.2. High-Frequency Pinching Artifacts12.3. A Robust Solution12.3.1. Smoothing Discontinuities12.3.2. Removing Pinches and Adding Detail12.4. Results12.5. Performance12.6. Caveats12.6.1. Forcing Convergence12.6.2. Tunable ParametersDistance AttenuationTriangle Attenuation12.7. Future Work12.8. References
13. Volumetric Light Scattering as a Post-Process
13.1. Introduction13.2. Crepuscular Rays13.3. Volumetric Light Scattering13.3.1. Controlling the Summation13.4. The Post-Process Pixel Shader13.5. Screen-Space Occlusion Methods13.5.1. The Occlusion Pre-Pass Method13.5.2. The Occlusion Stencil Method13.5.3. The Occlusion Contrast Method13.6. Caveats13.7. The Demo13.8. Extensions13.9. Summary13.10. References
III. Rendering
14. Advanced Techniques for Realistic Real-Time Skin Rendering
14.1. The Appearance of Skin14.1.1. Skin Surface Reflectance14.1.2. Skin Subsurface Reflectance14.2. An Overview of the Skin-Rendering System14.3. Specular Surface Reflectance14.3.1. Implementing a Physically Based Specular Reflectance Model for SkinRendering with a BRDFFresnel Reflectance for Rendering SkinFactoring BRDFs for Efficient EvaluationSpecular Reflectance from Skin Is WhiteVarying Specular Parameters over the Face14.4. Scattering Theory14.4.1. Diffusion Profiles14.4.2. Rendering with Diffusion Profiles14.4.3. The Shape of Diffusion Profiles14.4.4. A Sum-of-Gaussians Diffusion Profile14.4.5. Fitting Predicted or Measured Profiles14.4.6. Plotting Diffusion Profiles14.4.7. A Sum-of-Gaussians Fit for Skin14.5. Advanced Subsurface Scattering14.5.1. Texture-Space Diffusion14.5.2. Improved Texture-Space DiffusionMany BlursAdvantages of a Sum-of-Gaussians Diffusion ProfileCorrecting for UV DistortionThe Convolution ShaderOptional: Multiscale StretchingThe Accuracy of Stretch-Corrected Texture-Space DiffusionIncorporating Diffuse Color VariationPost-Scatter TexturingPre-Scatter TexturingCombining Pre-Scatter and Post-Scatter TexturingCompute Specular and Diffuse Light with the Same NormalsThe Final ShaderDealing with SeamsExact Energy Conservation14.5.3. Modified Translucent Shadow MapsMultiple Lights and Environment Lighting14.6. A Fast Bloom Filter14.7. Conclusion14.7.1. Future Work14.8. References
15. Playable Universal Capture
15.1. Introduction15.2. The Data Acquisition Pipeline15.3. Compression and Decompression of the Animated Textures15.3.1. Principal Component Analysis15.3.2. Compression15.3.3. Decompression15.3.4. Variable PCA15.3.5. Practical Considerations15.4. Sequencing Performances15.5. Conclusion15.6. References
16. Vegetation Procedural Animation and Shading in Crysis
16.1. Procedural Animation16.1.1. Implementation DetailsApproximating Sine WavesDetail BendingMain Bending16.2. Vegetation Shading16.2.1. Ambient Lighting16.2.2. Edge Smoothing16.2.3. Putting It All Together16.2.4. Implementation Details16.3. Conclusion16.4. References
17. Robust Multiple Specular Reflections and Refractions
17.1. Introduction17.2. Tracing Secondary Rays17.2.1. Generation of Layered Distance Maps17.2.2. Ray Tracing Layered Distance MapsLinear SearchAcceleration with Min-Max Distance ValuesRefinement by Secant Search17.3. Reflections and Refractions17.4. Results17.5. Conclusion17.6. References
18. Relaxed Cone Stepping for Relief Mapping
18.1. Introduction18.2. A Brief Review of Relief Mapping18.3. Cone Step Mapping18.4. Relaxed Cone Stepping18.4.1. Computing Relaxed Cone Maps18.4.2. Rendering with Relaxed Cone Maps18.5. Conclusion18.5.1. Further Reading18.6. References
19. Deferred Shading in Tabula Rasa
19.1. Introduction19.2. Some Background19.3. Forward Shading Support19.3.1. A Limited Feature Set19.3.2. One Effect, Multiple Techniques19.3.3. Light Prioritization19.4. Advanced Lighting Features19.4.1. Bidirectional Lighting19.4.2. Globe Mapping19.4.3. Box Lights19.4.4. Shadow MapsGlobal Shadow MapsLocal Shadow Maps19.4.5. Future Expansion19.5. Benefits of a Readable Depth and Normal Buffer19.5.1. Advanced Water and Refraction19.5.2. Resolution-Independent Edge Detection19.6. Caveats19.6.1. Material PropertiesChoose Properties WiselyEncapsulate and Hide MRT Data19.6.2. Precision19.7. Optimizations19.7.1. Efficient Light Volumes19.7.2. Stencil Masking19.7.3. Dynamic Branching19.8. Issues19.8.1. Alpha-Blended Geometry19.8.2. Memory Bandwidth19.8.3. Memory Management19.9. Results19.10. Conclusion19.11. References
20. GPU-Based Importance Sampling
20.1. Introduction20.2. Rendering Formulation20.2.1. Monte Carlo Quadrature20.2.2. Importance Sampling20.2.3. Sampling Material Functions20.3. Quasirandom Low-Discrepancy Sequences20.4. Mipmap Filtered Samples20.4.1. Mapping and Distortion20.5. Performance20.6. Conclusion20.7. Further Reading and References
IV. Image Effects
21. True Impostors
21.1. Introduction21.2. Algorithm and Implementation Details21.3. Results21.4. Conclusion21.5. References
22. Baking Normal Maps on the GPU
22.1. The Traditional Implementation22.1.1. Projection22.1.2. The Boundary Cage22.2. Acceleration Structures22.2.1. The Uniform Grid22.2.2. The 3D Digital Differential Analyzer22.3. Feeding the GPU22.3.1. Indexing Limitations22.3.2. Memory and Architectural Limitations22.4. Implementation22.4.1. Setup and Preprocessing22.4.2. The Single-Pass ImplementationStep 1Step 2Step 3Step 4Limitation22.4.3. The Multipass ImplementationLimitationTiled Rendering22.4.4. Antialiasing22.5. ResultsPossible Improvements22.6. Conclusion22.7. References
23. High-Speed, Off-Screen Particles
23.1. Motivation23.2. Off-Screen Rendering23.2.1. Off-Screen Depth Testing23.2.2. Acquiring DepthAcquiring Depth in DirectX 9Acquiring Depth in DirectX 1023.3. Downsampling Depth23.3.1. Point Sampling Depth23.3.2. Maximum of Depth Samples23.4. Depth Testing and Soft Particles23.5. Alpha Blending23.6. Mixed-Resolution Rendering23.6.1. Edge Detection23.6.2. Composing with Stenciling23.7. Results23.7.1. Image Quality23.7.2. Performance23.8. Conclusion23.9. References
24. The Importance of Being Linear
24.1. Introduction24.2. Light, Displays, and Color Spaces24.2.1. Problems with Digital Image Capture, Creation, and Display24.2.2. Digression: What Is Linear?24.2.3. Monitors Are Nonlinear, Renderers Are Linear24.3. The Symptoms24.3.1. Nonlinear Input Textures24.3.2. Mipmaps24.3.3. Illumination24.3.4. Two Wrongs Don’t Make a Right24.4. The Cure24.4.1. Input Images (Scans, Paintings, and Digital Photos)24.4.2. Output Images (Final Renders)24.4.3. Intermediate Color Buffers24.5. Conclusion24.6. Further Reading
25. Rendering Vector Art on the GPU
25.1. Introduction25.2. Quadratic Splines25.3. Cubic Splines25.3.1. Serpentine25.3.2. Loop25.3.3. Cusp25.3.4. Quadratic25.4. Triangulation25.5. Antialiasing25.6. Code25.7. Conclusion25.8. References
26. Object Detection by Color: Using the GPU for Real-Time Video Image Processing
26.1. Image Processing Abstracted26.2. Object Detection by Color26.2.1. Creating the Mask26.2.2. Finding the Centroid26.2.3. Compositing an Image over the Input Signal26.3. Conclusion26.4. Further Reading
27. Motion Blur as a Post-Processing Effect
27.1. Introduction27.2. Extracting Object Positions from the Depth Buffer27.3. Performing the Motion Blur27.4. Handling Dynamic Objects27.5. Masking Off Objects27.6. Additional Work27.7. Conclusion27.8. References
28. Practical Post-Process Depth of Field
28.1. Introduction28.2. Related Work28.2.1. Overview28.2.2. Specific Techniques28.3. Depth of Field28.4. Evolution of the Algorithm28.4.1. Initial Stochastic Approach28.4.2. The Scatter-as-Gather Approach28.4.3. The Blur Approach28.5. The Complete Algorithm28.5.1. Depth Information28.5.2. Variable-Width Blur28.5.3. Circle of Confusion Radius28.5.4. First-Person Weapon Considerations28.5.5. The Complete Shader Listing28.6. Conclusion28.7. Limitations and Future Work28.8. References
V. Physics Simulation
29. Real-Time Rigid Body Simulation on GPUs
29.1. Introduction29.1.1. Translation29.1.2. Rotation29.1.3. Shape Representation29.1.4. Collision Detection29.1.5. Collision Reaction29.2. Rigid Body Simulation on the GPU29.2.1. Overview29.2.2. The Data Structure29.2.3. Step 1: Computation of Particle Values29.2.4. Step 2: Grid Generation29.2.5. Step 3: Collision Detection and Reaction29.2.6. Step 4: Computation of Momenta29.2.7. Step 5: Computation of Position and Quaternion29.2.8. Rendering29.2.9. Performance29.3. Applications29.3.1. Granular Materials29.3.2. Fluids29.3.3. Coupling29.4. Conclusion29.5. Appendix29.6. References
30. Real-Time Simulation and Rendering of 3D Fluids
30.1. Introduction30.2. Simulation30.2.1. Background30.2.2. Equations of Fluid Motion30.2.3. Solving for VelocityImproving Detail30.2.4. Solid-Fluid InteractionDynamic ObstaclesVoxelizationInside-Outside VoxelizationVelocity VoxelizationOptimizing Voxelization30.2.5. Smoke30.2.6. Fire30.2.7. Water30.2.8. Performance Considerations30.2.9. Storage30.2.10. Numerical Issues30.3. Rendering30.3.1. Volume RenderingVolume Ray CastingCompositingClippingFilteringOff-Screen Ray MarchingFire30.3.2. Rendering LiquidsRefraction30.4. Conclusion30.5. References
31. Fast N-Body Simulation with CUDA
31.1. Introduction31.2. All-Pairs N-Body Simulation31.3. A CUDA Implementation of the All-Pairs N-Body Algorithm31.3.1. Body-Body Force Calculation31.3.2. Tile Calculation31.3.3. Clustering Tiles into Thread Blocks31.3.4. Defining a Grid of Thread Blocks31.4. Performance Results31.4.1. OptimizationPerformance Increase with Loop UnrollingPerformance Increase as Block Size VariesImproving Performance for Small N31.4.2. Analysis of Performance Results31.5. Previous Methods Using GPUs for N-Body Simulation31.6. Hierarchical N-Body Methods31.7. Conclusion31.8. References
32. Broad-Phase Collision Detection with CUDA
32.1. Broad-Phase Algorithms32.1.1. Sort and Sweep32.1.2. Spatial Subdivision32.1.3. Parallel Spatial Subdivision32.2. A CUDA Implementation of Spatial Subdivision32.2.1. Initialization32.2.2. Constructing the Cell ID Array32.2.3. Sorting the Cell ID ArrayThe Radix Sort AlgorithmThe Parallel Radix Sort AlgorithmPhase 1: Setup and TabulationPhase 2: Radix SummationPhase 3: Reordering32.2.4. Creating the Collision Cell List32.2.5. Traversing the Collision Cell List32.3. Performance Results32.4. Conclusion32.5. References
33. LCP Algorithms for Collision Detection Using CUDA
33.1. Parallel Processing33.2. The Physics Pipeline33.3. Determining Contact Points33.3.1. Continuum Methods33.3.2. Discrete Methods (Coherence Based)33.3.3. Resolving Contact Points33.4. Mathematical Optimization33.4.1. Linear Programming33.4.2. The Linear Complementarity Problem33.4.3. Quadratic Programming33.5. The Convex Distance Calculation33.5.1. Lemke’s Algorithm for Solving the LCP33.6. The Parallel LCP Solution Using CUDA33.6.1. Implementation of the SolverSelecting the Pivot ElementSolving for the Selected Equation Variable33.7. Results33.8. References
34. Signed Distance Fields Using Single-Pass GPU Scan Conversion of Tetrahedra
34.1. Introduction34.1.1. Overview of Signed Distance Fields34.1.2. Overview of Our Method34.2. Leaking Artifacts in Scan Methods34.2.1. The Plane Test34.2.2. How a Bounding Volume Is Constructed34.2.3. Folds in the Polygonal Model34.3. Our Tetrahedra GPU Scan Method34.3.1. Computing the Shell34.3.2. Computing the Cross Section of a Tetrahedron34.3.3. Computing Signed Distance Using Angle-Weighted Pseudonormals34.4. Results34.5. Conclusion34.6. Future Work34.6.1. Improvements to the Algorithm34.6.2. Improvements to Our Implementation34.7. Further Reading34.8. References
VI. GPU Computing
35. Fast Virus Signature Matching on the GPU
35.1. Introduction35.2. Pattern Matching35.2.1. A Data-Scanning Library35.3. The GPU Implementation35.4. Results35.5. Conclusions and Future Work35.6. References
36. AES Encryption and Decryption on the GPU
36.1. New Functions for Integer Stream Processing36.1.1. Transform Feedback Mode36.1.2. GPU Program Extensions36.2. An Overview of the AES Algorithm36.3. The AES Implementation on the GPU36.3.1. Input/Output and the State36.3.2. Initialization36.3.3. RoundsThe SubBytes OperationThe ShiftRows OperationThe MixColumns OperationThe AddRoundKey Operation36.3.4. The Final Round36.4. Performance36.4.1. Variable Batch Size36.4.2. Comparison to CPU-Based Encryption36.5. Considerations for Parallelism36.5.1. Block-Cipher Modes of OperationElectronic Code Book ModeCipher-Block Chaining Mode36.5.2. Modes for Parallel ProcessingDecryption in CBC ModeCounter Mode36.6. Conclusion and Future Work36.7. References
37. Efficient Random Number Generation and Application Using CUDA
37.1. Monte Carlo Simulations37.2. Random Number Generators37.2.1. Introduction37.2.2. Uniform-to-Gaussian Conversion GeneratorAvailable Uniform PRNGsLinear Congruential GeneratorMultiple Recursive GeneratorLagged Fibonacci GeneratorMersenne TwisterCombined Tausworthe GeneratorA Hybrid GeneratorKISS37.2.3. Types of Gaussian TransformsThe Ziggurat MethodThe Polar MethodThe Box-Muller Transform37.2.4. The Wallace Gaussian GeneratorPermuting the PoolInitializing the Pool37.2.5. Integrating the Wallace Gaussian Generator into a Simulation37.3. Example Applications37.3.1. Asian Option37.3.2. Variant on a Lookback Option37.3.3. Results37.4. Conclusion37.5. References
38. Imaging Earth’s Subsurface Using CUDA
38.1. Introduction38.2. Seismic Data38.3. Seismic Processing38.3.1. Wave Propagation38.3.2. Seismic Migration Using the SRMIP Algorithm38.4. The GPU Implementation38.4.1. GPU/CPU Communication38.4.2. The CUDA Implementation38.4.3. The Wave Propagation Kernel38.5. Performance38.6. Conclusion38.7. References
39. Parallel Prefix Sum (Scan) with CUDA
39.1. Introduction39.1.1. Sequential Scan and Work Efficiency39.2. Implementation39.2.1. A Naive Parallel Scan39.2.2. A Work-Efficient Parallel Scan39.2.3. Avoiding Bank Conflicts39.2.4. Arrays of Arbitrary Size39.2.5. Further Optimization and Performance Results39.2.6. The Advantages of CUDA over the OpenGL Implementation39.3. Applications of Scan39.3.1. Stream Compaction39.3.2. Summed-Area Tables39.3.3. Radix SortStep 1: Radix Sort ChunksStep 2: Merge Sorted Chunks39.3.4. Previous Work39.4. Conclusion39.5. References
40. Incremental Computation of the Gaussian
40.1. Introduction and Related Work40.2. Polynomial Forward Differencing40.3. The Incremental Gaussian Algorithm40.4. Error Analysis40.5. Performance40.6. Conclusion40.7. References
41. Using the Geometry Shader for Compact and Variable-Length GPU Feedback
41.1. Introduction41.2. Why Use the Geometry Shader?41.3. Dynamic Output with the Geometry Shader41.4. Algorithms and Applications41.4.1. Building Histograms41.4.2. CompressorsVariable-Length Output41.4.3. The Hough Transform41.4.4. Corner Detection41.5. Benefits: GPU Locality and SLI41.6. Performance and Limits41.6.1. Guidelines41.6.2. Performance of the Hough Map Maxima Detection41.7. Conclusion41.8. References
Addison-Wesley Warranty on the DVD
Addison-Wesley Warranty on the DVDNVIDIA Statement on the SoftwareNo WarrantyLimitation of LiabilityDVD System Requirements
Inside Back Cover
GeometryLight and ShadowsRenderingImage EffectsPhysics SimulationGPU Computing

Content preview from GPU Gems 3

Part VI. GPU Computing

Figure .

In GPU Gems 3, we continue to showcase work that uses graphics hardware for nongraphics computation. As each new generation provides significantly greater computing power and programmability, GPUs are increasingly attractive targets for general-purpose computation, or what is commonly called GPGPU or GPU Computing. As a result, researchers and developers in academia and industry continue to develop new GPU algorithms for tasks such as sorting, database operations, image processing, and linear algebra. In many cases, the principal motivation for using the GPU is the prospect of high performance at a relatively low ...

Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.

Read now

Unlock full access

More than 5,000 organizations count on O’Reilly

O’Reilly covers everything we've got, with content to help us build a world-class technology community, upgrade the capabilities and competencies of our teams, and improve overall team performance as well as their engagement.

Julian F.

Head of Cybersecurity

I wanted to learn C and C++, but it didn't click for me until I picked up an O'Reilly book. When I went on the O’Reilly platform, I was astonished to find all the books there, plus live events and sandboxes so you could play around with the technology.

Addison B.

Field Engineer

I’ve been on the O’Reilly platform for more than eight years. I use a couple of learning platforms, but I'm on O'Reilly more than anybody else. When you're there, you start learning. I'm never disappointed.

Amir M.

Data Platform Tech Lead

I'm always learning. So when I got on to O'Reilly, I was like a kid in a candy store. There are playlists. There are answers. There's on-demand training. It's worth its weight in gold, in terms of what it allows me to do.

Mark W.

Embedded Software Engineer

Publisher Resources

ISBN: 9780321545428Purchase book

Cloud Computing

Data Engineering

Data Science

AI & ML

Programming Languages

Software Architecture

IT/Ops

Security

Design

Business

Soft Skills

GPU Gems 3

by Hubert Nguyen

Part VI. GPU Computing

Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.