In Situ Analysis and Visualization of Extreme-Scale Particle Simulations

In situ analysis has emerged as a dominant paradigm for performing scalable visual analysis of extreme-scale computational simulation data. Compared to the traditional post hoc analysis pipeline where data is first stored into disks and then analyzed offline, in situ analysis processes data at the time its generation in the supercomputers so that the slow and expensive disk I/O is minimized. In this work, we present a new in situ visual analysis pipeline for the extreme-scale multi-phase flow simulation MFiX-Exa and demonstrate how the pipeline can be used to process large particle fields in situ and produce informative visualizations of the data features. We deploy our analysis pipeline on Oak Ridge's Summit supercomputer to study its in situ applicability and usefulness.