One of the most difficult problems in volume rendering hardware architectures is to provide an efficient hardware integration of algorithmic optimization techniques like, space leaping and early ray termination. An architecture, which solves this problem, is presented in this paper. It is based on a multithreaded approach that overcomes branch hazards caused by these algorithmic optimization techniques. Additionally, a new memory architecture suitable for the multithreaded processing of rays is presented. Simulations show efficiencies of up to 97% (only 3% idle time) for the full system. A frame rate of 132 Hz for a data set of 256 256 128 and up to 4 Hz for 1024 3 voxels has been found by simulation.
Dieser Eintrag ist Teil der Universitätsbibliographie.
Search Authors in