This Closes https://github.com/cadop/dhart/issues/29 . The issue was partly due to scheduling of openmp and resolved by changing dynamic allocation size. The streaming method is only beneficial for coherent rays, and just because direction is the same does not mean the rays are coherent, they also have to be from the almost same origin.
Modifies:
Occlusions method for the one direction multiple origins case.
Chunk sizes for all occlusion tests to be 128 in openmp: #pragma omp parallel for if(use_parallel) schedule(dynamic, 128)
Adds:
Unittest to ensure multi origin rays is returning correct order
This Closes https://github.com/cadop/dhart/issues/29 . The issue was partly due to scheduling of openmp and resolved by changing dynamic allocation size. The streaming method is only beneficial for coherent rays, and just because direction is the same does not mean the rays are coherent, they also have to be from the almost same origin.
Modifies:
Occlusionsmethod for the one direction multiple origins case.128in openmp:#pragma omp parallel for if(use_parallel) schedule(dynamic, 128)Adds: