SC²S Colloquium - May 8, 2017

From Sccswiki
Jump to navigation Jump to search
Date: May 8, 2017
Room: 01.09.014
Time: 10:30 am, s.t.

Daniel Endress: Generic Output for Mesh-Based Supercomputing Applications

High-performance computing is an important component for various scientific researches. Many physical, seismological, or astrological simulations would not be possible without the use of super-computing technology. In order to ensure the best performance, not only the simulation needs to be optimized for a parallel environment but the output process as well. This optimization of the output mechanism is not trivial and involves various difficulties. The aim of this thesis is to provide a generic framework that can be used by many different HPC applications for generating their output. It handles the said difficulties and provides an easy to use API to be utilized by the simulations. Due to its abstract design it is applicable to most mesh-based computations and therefore prevents an unnecessary reinvention of output writing techniques by each simulation on their own. The library has been successfully integrated into three projects as part of this thesis. In doing so, it proved its functionality and showed good performance and scalability. Therefore, this library offers new possibilities while still maintaining optimal performance.