Spatial Partitioning For Hierarchical Radiosity
Spatial Partitioning For Hierarchical Radiosity,
Grafiktag 2000. Workshop über Trends und
Höhepunkte der Graphischen Datenverarbeitung,
im Rahmen der 30.
Jahrestagung der GI,
Sep 19, 2000, Berlin, Germany.
This work resides around the interface of two
disciplines in computer science: computer graphics (global illumination)
and parallel computing (dynamic partitioning).
On the one hand an efficient algorithm for global illumination simulation
- hierarchical radiosity - is examined with respect to its capability
of being parallelized. On the other hand we develop a general
tool T for the dynamic partitioning of spatially mapped tasks, that is
furthermore analyzed theoretically and experimentally.
The hierarchical radiosity algorithm is a special instance
of algorithms that can be formulated as a collection
of spatially mapped tasks. As a proof of practicability of our
tool T we apply the tool to the hierarchical radiosity algorithm
and get useful speedup values. Algorithms of other scientific
computing areas apparently can benefit the same way from our tool,
even though studying this is beyond the scope of this paper.