Extreme-Scale In Situ Visualization of Turbulent Flows on IBM Blue Gene/Q JUQUEEN

摘要

Extracting and analyzing detailed information from large simulations is of crucial importance for science. However, with the increasing problem size of current simulations, the process of visualizing and understanding big simulation raw data becomes more difficult and needs additional effort. More precisely, the gap between compute and I/O performance is widening with current supercomputers. Thus, the classical approach of visualizing simulation results in a post-processing step is limited or even impossible for extreme-scale scenarios. One promising technique to overcome this issue is in situ visualization, which visualizes and analyzes simulation data during simulation runtime. Within this work, in situ visualization using VisIt/Libsim has been added to the CIAO code framework for interactive- and batch-mode visualization on JUQUEEN, an IBM Blue Gene/Q system with 458 752 cores. Full-system runs are demonstrated and early results of performance measurements of an extreme-scale multiphase case are discussed.