The Render Cache [1,2] allows the interactive display of very large scenes, rendered with complex global illumination models, by decoupling camera movement from the costly scene sampling process. In this paper, the distributed execution of the individual components of the Render Cache on a PC cluster is shown to be a viable alternative to the shared memory implementation.As the processing power of an entire node can be dedicated to a single component, more advanced algorithms may be examined. Modular functional units also lead to increased flexibility, useful in research as well as industrial applications.We introduce a new strategy for view-driven scene sampling, as well as support for multiple camera viewpoints generated from the same cache. Stereo display and a CAVE multi-camera setup have been implemented.The use of the highly portable and inter-operable CORBA networking API simplifies the integration of most existing pixel-based renderers. So far, three renderers (C++ and Java) have been adapted to function within our framework.
[1]
Donald P. Greenberg,et al.
Interactive global illumination in dynamic scenes
,
2002,
SIGGRAPH.
[2]
George Drettakis,et al.
Enhancing and Optimizing the Render Cache
,
2002,
Rendering Techniques.
[3]
George Drettakis,et al.
Interactive Rendering using the Render Cache
,
1999,
Rendering Techniques.
[4]
Philipp Slusallek,et al.
SaarCOR: a hardware architecture for ray tracing
,
2002,
HWWS '02.
[5]
Donald P. Greenberg,et al.
Combining edges and points for interactive high-quality rendering
,
2003,
ACM Trans. Graph..
[6]
Erik Reinhard,et al.
Parallel point reprojection
,
2001,
Proceedings IEEE 2001 Symposium on Parallel and Large-Data Visualization and Graphics (Cat. No.01EX520).