SC19 Proceedings

The International Conference for High Performance Computing, Networking, Storage, and Analysis

Local-Global Merge Tree Computation with Local Exchanges

Authors: Arnur Nigmetov (Graz University of Technology), Dmitriy Morozov (Lawrence Berkeley National Laboratory)

Abstract: A merge tree is a topological summary of a real-valued function on a graph. Merge trees can be used to find stable features in the data, report the number of connected components above any threshold, or compute other topological descriptors. A local-global merge tree provides a way of distributing a merge tree among multiple processors so that queries can be performed with minimal communication. While this makes them efficient in massively parallel setting, the only known algorithm for computing a local-global merge tree involves global reduction.

Motivated by applications in cosmological simulations, we consider a restricted version of the problem: we compute a local-global tree down to a threshold fixed by the user. We describe two algorithms for computing such a tree via only local exchanges between processors. We present a number of experiments that show the advantage of our method on different simulations.

Back to Technical Papers Archive Listing