Para06 - Minisymposium - MS16

Organizers and Chair persons:

As high-performance computing systems become more complex crunching on and generating terabytes or petabytes of data, it becomes more and more important to be able to visualize the results and utilize state-of-the-art tools to develop and optimize applications on modern HPC systems, including large parallel clusters and grids.

This minisymposium hightlights some of the recent developments in this area. The symposium is dived into two parts where the first part focuses on visualization environments and the second on HPC tools.

In the first part we start off with an brief overview of HPC environments including an overview of the topics to be presented related to visualization. Stødle et. al. will then be presenting their work on using Shared Windows to support collaboration, monitoring and debugging on distributed displays including display walls driven by clusters. In the third talk, Meyer will present a stratetgy for efficiently handling computation of large, fixed data sets. This work originates from the need to adapt an existing load balancing scheme for a server which animates visualizations of geological data sets in real-time to suit a multi-user environment. Nagel will illustrate the similarities and differences between Scientific Visualization and Information Visualization. Natvig and Elster may present a cute technique for fast rendering of large datasets. Saltvik ends the visualizatioin part of this minisymposium by presenting a visualization applications that uses parallel techniques to achieve the performance needed to get good performance and realism in simualting snow.

The second part of this minisymposium will focus on parallelization tools, including enhancements to MPI (second & thrid talk), a look at issues with benchmarking MPI collective operations (possible 4th talk) and applications with floating point numbers. The last presentation will give an overview of some of the tools developed as part of collarborations between NTNU and CERN.