Numerical simulation has become extremely important in science since it ideally supports classical approaches like theoretical considerations and experimental measurements. For this reason, simulations are largely used in the DLR to better understand complex phenomena. This includes areas like computational fluid dynamics, structural mechanics simulations or the shape optimization of transport vehicles. Accurate numerical simulations usually involve a high computational complexity which leads to enormous costs. It is therefore important to efficiently use the computer systems on which these simulations are performed.
Our focus is on efficient numerical simulations. For this purpose, we investigate the parallel scaling behavior of algorithms and data structures for massively parallel computers. Furthermore, we develop software that is optimized for GPU- and many-core CPU systems. Finally, we also consider an efficient management of large datasets. These datasets do not only result from computer simulations but are also obtained in the use or the monitoring of aircrafts and space vehicles.
The research of the department for High-Performance Computing focuses on:
Parallel Numerical Mathematics
Modeling and shape optimization
Software for future architectures
Efficient treatment of large datasets and Big Data