A Simple, General, and GPU Friendly Method for Computing Dual Mesh and Iso-Surfaces of Adaptive Mesh Refinement (AMR) Data
share
We propose a novel approach to extracting crack-free iso-surfaces from Structured AMR data that is more general than previous techniques, is trivially simple to implement, requires no information other than the list of AMR cells, and works, in particular, for different AMR formats including octree AMR, block-structured AMR with arbitrary level differences at level boundaries, and AMR data that consist of individual cells without any existing grid structure. We describe both the technique itself and a CUDA-based GPU implementation of this technique, and evaluate it on several non-trivial AMR data sets.
READ FULL TEXT