Session Details
MS03-1: Challenges of Current and Emerging Applications for High Performance Computing (Ganzes Minisymposium anzeigen)
Thursday, 12. October 2017; 10:30 - 12:30 Uhr in Raum 7.12
Sitzungsleitung: Ralf Schneider
10:30
An HPC technology review with respect to large scale engineering applications
Ralf Schneider (High Performance Computing Center Stuttgart)
Kurzfassung:
An overview about the current state of the art in HPC-systems and technology will be given. Especially the current trends in accelerators, vector and many-core architectures, resulting from the stagnating scalar per core performance of classical CPUs will be addressed. In view of large scale engineering applications from the fields of structural mechanics and fluid dynamics the current and emerging bottlenecks as well as the technological and intellectual challenges will be discussed.
An HPC technology review with respect to large scale engineering applications
Ralf Schneider (High Performance Computing Center Stuttgart)
Kurzfassung:
An overview about the current state of the art in HPC-systems and technology will be given. Especially the current trends in accelerators, vector and many-core architectures, resulting from the stagnating scalar per core performance of classical CPUs will be addressed. In view of large scale engineering applications from the fields of structural mechanics and fluid dynamics the current and emerging bottlenecks as well as the technological and intellectual challenges will be discussed.
10:50
Numerical Simulation of Flow with Volume Condensation during Accident in Containment of Nuclear Power Plant
Jing Zhang (High Performance Computing Center Stuttgart)
Kurzfassung:
One severe accident scenario is a leak in the primary circuit of a Pressurized Water Reactor (PWR), resulting in hydrogen and steam injection into the containment. Because of the influence of steam condensation on the gas mixing, the wall and volume condensation phenomena are of interest for the safety considerations. This presentation shows the simulation results of two-phase flow using the developed volume condensation model in the containment. Simulations were performed on the HPC systems.
Numerical Simulation of Flow with Volume Condensation during Accident in Containment of Nuclear Power Plant
Jing Zhang (High Performance Computing Center Stuttgart)
Kurzfassung:
One severe accident scenario is a leak in the primary circuit of a Pressurized Water Reactor (PWR), resulting in hydrogen and steam injection into the containment. Because of the influence of steam condensation on the gas mixing, the wall and volume condensation phenomena are of interest for the safety considerations. This presentation shows the simulation results of two-phase flow using the developed volume condensation model in the containment. Simulations were performed on the HPC systems.
11:10
Modelling the neuromuscular system using HPC systems
Thomas Klotz (University of Stuttgart), Nehzat Emamy (University of Stuttgart), Thomas Ertl (University of Stuttgart), Dominik Göddeke (University of Stuttgart), Aaron Krämer (University of Stuttgart), Michael Krone (University of Stuttgart), Benjamin Maier (University of Stuttgart), Miriam Mehl (University of Stuttgart), Tobias Rau (University of Stuttgart), Oliver Röhrle (University of Stuttgart)
Kurzfassung:
Modelling the neuromusclular system is challenging due to its high compexlity and variability. Formulating models that account for a realistic biophysical motivated activation process leads to computational expensive multi-scale simulations, which, in turn limits on normal compute environments the model detail and model size due to its computational complexity. We aim to overcome these limitations by using massively parallel HPC clusters.
Modelling the neuromuscular system using HPC systems
Thomas Klotz (University of Stuttgart), Nehzat Emamy (University of Stuttgart), Thomas Ertl (University of Stuttgart), Dominik Göddeke (University of Stuttgart), Aaron Krämer (University of Stuttgart), Michael Krone (University of Stuttgart), Benjamin Maier (University of Stuttgart), Miriam Mehl (University of Stuttgart), Tobias Rau (University of Stuttgart), Oliver Röhrle (University of Stuttgart)
Kurzfassung:
Modelling the neuromusclular system is challenging due to its high compexlity and variability. Formulating models that account for a realistic biophysical motivated activation process leads to computational expensive multi-scale simulations, which, in turn limits on normal compute environments the model detail and model size due to its computational complexity. We aim to overcome these limitations by using massively parallel HPC clusters.
11:30
Load-balance strategies for CFD-codes on HPC systems
Philipp Offenhäuser (High Performance Computing Center Stuttgart)
Kurzfassung:
Today's HPC systems generate their performance by facilitating hundreds of thousands of cores. In order to use this computing power efficiently, the computational effort must be distributed evenly across all cores. Techniques for distributing the simulation initially are well-known. Based on numerical and physical phenomena additional computational effort may occur locally, at run-time. Techniques are presented which recognize these additional loads and redistribute the simulation evenly.
Load-balance strategies for CFD-codes on HPC systems
Philipp Offenhäuser (High Performance Computing Center Stuttgart)
Kurzfassung:
Today's HPC systems generate their performance by facilitating hundreds of thousands of cores. In order to use this computing power efficiently, the computational effort must be distributed evenly across all cores. Techniques for distributing the simulation initially are well-known. Based on numerical and physical phenomena additional computational effort may occur locally, at run-time. Techniques are presented which recognize these additional loads and redistribute the simulation evenly.
11:50
Eigenmodes for nonlinear operators
Uwe Küster (High Performance Computing Center Stuttgart), Ralf Schneider (High Performance Computing Center Stuttgart), Andreas Ruopp (High Performance Computing Center Stuttgart)
Kurzfassung:
Even if spectral analysis seems to be restricted to linear operators, it turns out that it is also applicable to nonlinear operators by turning them linear by embedding these in a much larger space and analyzing the Koopman operator. Spectral analysis will be possible but in an infinite dimensional space. We show a numerical approach, which allows to separate different parts of instationary data in a time vanishing part and a remaining part. We discuss implications for calculation and IO.
Eigenmodes for nonlinear operators
Uwe Küster (High Performance Computing Center Stuttgart), Ralf Schneider (High Performance Computing Center Stuttgart), Andreas Ruopp (High Performance Computing Center Stuttgart)
Kurzfassung:
Even if spectral analysis seems to be restricted to linear operators, it turns out that it is also applicable to nonlinear operators by turning them linear by embedding these in a much larger space and analyzing the Koopman operator. Spectral analysis will be possible but in an infinite dimensional space. We show a numerical approach, which allows to separate different parts of instationary data in a time vanishing part and a remaining part. We discuss implications for calculation and IO.