Data Parallel Hypersweeps for in Situ Topological Analysis

Abstract

The contour tree is a tool for understanding the topological structure of a scalar field. Recent work has built efficient contour tree algorithms for shared memory parallel computation, driven by the need to analyze large data sets in situ while the simulation is running. Unfortunately, methods for using the contour tree for practical data analysis are still primarily serial, including single isocontour extraction, branch decomposition and simplification. We report data parallel methods for these tasks using a data structure called the hyperstructure and a general purpose approach called a hypersweep. We implement and integrate these methods with a Cinema database that stores features as depth images and with a web server that reconstructs the features for direct visualization.

Distributed Hierarchical Contour Trees

“…In each iteration, monotone chains of superarcs transfer as hyperarcs to the contour tree -i.e. a specialised form of branch decomposition [24] that is related to rake-reduce parallel tree algorithms [16]. In this way, a hyperstructure is built up that allows logarithmic search in the contour tree for efficient parallel processing.…”

“…In subsequent work [16], we then showed that the hyperstructure was also related to the Euler Tour [4], which allows computations over trees to be collapsed into prefix sum operations. By exploiting this, we demonstrated efficient shared-memory parallel computation of geometric measures, simplification, branch decomposition and isocontour extraction.…”

“…This summation is a modified subtree size computation, and can be performed efficiently in parallel with an Euler Tour or a hypersweep [16]. In practice, rather than split superarcs up, we define a superarc to be necessary for the boundary tree if none of the regular arcs on it are in the interior forest, and treat the supernodes at both ends as necessary as well.…”