Scalable In Situ Lagrangian Flow Map Extraction: Demonstrating the Viability of a Communication-Free Model

• DOI: 10.48550/arxiv.2004.02003
• Publication Date: 2020-01-01
• Authors: Sudhanshu Sane; Abhishek Yenpure; Roxana Bujack; Matthew Larsen; Kenneth Moreland; Christoph Garth; Hank Childs

We introduce and evaluate a new algorithm for the in situ extraction of Lagrangian flow maps, which we call Boundary Termination Optimization (BTO). Our approach is communication-free model, requiring no message passing or synchronization between processes, improving scalability, thereby reducing overall execution time alleviating encumbrance placed on simulation codes from processing. terminate particle integration at node boundaries store only subset map that would have been extracted by communicating particles across nodes, thus introducing an accuracy-performance tradeoff. run experiments with as many 2048 GPUs multiple data sets. For experiment configurations consider, our findings demonstrate technique saves much 2x to 4x situ, while staying nearly accurate quantitatively qualitatively previous work. Most significantly, this study establishes viability approaching using models future.