H. Du, N. Nabrizi, and N. Bagherzadeh (USA)
Parallel Ray Tracing, SIMD
We study efficiency issues to implement a parallel ray tracing model onto SIMD architectures. We investigate various mechanisms to synchronize the parallel threads existing in the parallel ray-tracing algorithm. We also exploit data and ray coherence in order to efficiently utilize the available memory bandwidth. We have implemented our findings on a reconfigurable SIMD engine, MorphoSys. According to our encouraging results these findings prove decisive in implementing high performance SIMD ray-tracing systems on flexible platforms.
Important Links:
Go Back
