Code Optimization and Stabilization for a High-Resolution Terrain Generation Application
TimeTuesday, July 246:30pm - 8:30pm
DescriptionAccess to submeter resolution commercial imagery through NSF-sponsored research centers such as the Polar Geospatial Center (PGC), has the potential to revolutionize Earth observation. To do so, there must be an efficient system for producing and distributing digital elevation models, DEMs. The Surface Extraction from TIN based Search-space Minimization (SETSM) software from The Ohio State University (OSU) provides fully-automated DEM extraction using such imagery to produce DEM-based terrain maps.
In this poster, we will present code improvements that Ohio Supercomputer Center and OSU researchers have made to SETSM to make this capability widely available. The improvements include software engineering efforts undertaken to improve code portability and stability prior to making SETSM available on a public repository. Several optimizations have been incorporated to increase performance of important loop nests such as vectorization and code refactoring to reduce redundant operations. The current release of the SETSM uses OpenMP, and we are in the process of refining a recent hybrid MPI/OpenMP parallel implementation which will be made available in a future release. Many of these improvements have been used in the ArcticDEM Production project on Blue Waters to construct the first high-resolution (2m), high-quality (<5m uncertainty), and openly distributed DEM set of all Arctic land masses. We will discuss these efforts as well as some of the challenges of moving a rapidly evolving research code into an open source project.