Register
Speaker: Alex Razoumov (SFU)
Abstract: As numerical sim
ulations continue to grow in size and complexity\, the I/O bottleneck make
s it impossible to save all generated data for later analysis. In-situ tec
hniques address this challenge by running analysis and visualization in re
al time\, while the data are still in memory on the HPC system running the
simulation\, eliminating the need to write everything to disk. Several in
-situ visualization frameworks exist\, including ParaView'\;s Catalyst\
, VisIt'\;s LibSim\, Ascent\, and SENSEI.
In this hands-on works
hop\, we will focus on Catalyst2\, which allows you to analyze and visuali
ze your simulation data using familiar ParaView pipelines. Catalyst2 provi
des an API for describing and passing data arrays -- computational meshes
and fields -- from your simulation to the library\, which then converts th
ese arrays into the appropriate VTK data structures. This happens without
requiring the user to understand the underlying VTK data model and without
duplicating data in memory. This framework is designed to scale to very l
arge datasets and to thousands of CPU cores via MPI.
The goal of th
is workshop is to instrument a simple C code with the Catalyst2 library an
d produce in-situ outputs\, including data extracts and rendered images\,
using ParaView'\;s Python scripting.
Prerequisites: To participa
te in the hands-on exercises\, please install ParaView on your computer. Y
ou can download it from the official website. Familiarity with ParaVi
ew basics -- creating a visualization pipeline\, applying filters\, and sc
ripting -- is assumed and covered in earlier sessions. Some knowledge of C
would be helpful\, but we'\;ll take things step by step.