| | |
Summary: TigerSHARK: A Hardware Accelerated Ray-tracing Engine
Greg Humphreys
Princeton University
humper@cs.princeton.edu
C. Scott Ananian
Princeton University
cananian@ee.princeton.edu
Abstract The current state of the art in graphics rendering algo-
rithms and hardware is surveyed, and it is shown that ray-tracing,
despite larger computational requirements than conventional algo-
rithms, is more amenable to massive parallelism. The TigerSHARK
ray-tracing architecture is then presented as an extremely cost-effective
means to exploit this fundamental parallelism. A competing approach
is analyzed, and the TigerSHARK SIMD-variant architecture shown
superior due to its use of low-cost high-performance reprogrammable
computing elements and specialized architecture. Prototype hardware
is briefly described. The system produces extremely high-quality out-
put, efficiently using low-cost hardware to rival the rendering speeds of
systems three orders of magnitude more expensive.
INTRODUCTION
|
