1 / 11

Integrating GPGPU Functionality into Scene Graphs

Jens Orthmann , Maik Keller, Andreas Kolb Computer Graphics and Multimedia Systems University of Siegen. Integrating GPGPU Functionality into Scene Graphs. Motivation. Evolution of generic GPGPU languages Heterogeneous architectures (OpenGL + CUDA)

bono
Download Presentation

Integrating GPGPU Functionality into Scene Graphs

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Jens Orthmann, Maik Keller, Andreas Kolb Computer Graphics and Multimedia Systems University of Siegen Integrating GPGPU Functionality into Scene Graphs

  2. Motivation • Evolution of generic GPGPU languages • Heterogeneous architectures (OpenGL + CUDA) • Scene graphs are the most frequently used graphics architectures • General purpose functionality for scene graphs ( Physics, Simulations, Ray Tracing, Image Processing, Visualization ) NVSG by NVidia VolumeViz by Mercury XIP by Siemens

  3. Contexts • A scene graph is populated with different kinds of rendering resources. (textures, geometry, etc.) • Developers need to map these resources to the CUDA or OpenCL context • Interoperability should be invisible for developers. • The interface should be small. Rendering Computation

  4. Graph Integration • How to integrate computations into the graph? • How to identify resources? • How to manage dynamic structures? ?

  5. Approaches Callback Approach Engine Approach Subtraversal Approach • SmallInterface • Fixed References • No Intracommunication • Intracommunication • Multiple resources • Hidden dependencies • Fixed References • No references • Intracommunication • Multiple resources • Visible dependencies • Requires traversals • Naming scheme ’’Input’’ ’’Input’’ ’’Output’’

  6. Subtraversal Approach Single Context Shared Context Multiple Contexts

  7. Resource Gathering • Resources are passed on to the computation modules • Exchange current computation context

  8. Graph Evaluation • Resources are mapped during the computation • Abstract view of resources within computations

  9. Resource Mapping • Intermixing rendering with computations • Memory mapping between contexts is expensive

  10. Results osgCompute: • CUDA (2.3) integration into OpenSceneGraph • Useable without the Scenegraph structure (similar to the Thrust API) • Support for all platforms: Linux, Mac, Windows • Framework, documentation and examples are available under the LGPL: http://www.cg.informatik.uni-siegen.de/svt/osgcompute Current Research:

  11. This work is partially funded by the Siegener Graduate School “Development of Integral Heterosensor Architectures for the n-Dimensional (Bio)chemical Analysis”. • Thanks to Robert Osfield and Art Tevs for their support Thank you for your attention!

More Related