"Unify" single node and distributed contour tree filters
Currently, we have ContourTreeUniformAugmented
, which computes a contour tree on a single (multi-threaded) compute node and ContourTreeUniformDistributed
, which computes a distributed representation of the contour tree using DIY/MPI. ContourTreeUniformDistributed
throws an exception if called for a single block. We should explore the possibility of "unifying" both filters or at least coming up with an approach that makes it possible to call the ContourTreeUniformDistributed
on a data set consisting of only one partition. E.g., what should be returned in that edge case? What makes most sense for analysis. Addressing this issue should be part of a general clean-up of the contour tree API.