PDExa: Optimized software methods for solving partial differential equations on exascale supercomputers
Project overview
The project PDExa, funded by the German Federal Ministry of Education and Research (BMBF) through the initiative SCALEXA, project id 16ME0637K, develops novel software method for efficiently solving partial differential equations on supercomputer scale. The project combines the expertise of five partners at Technical University of Munich, the University of Augsburg, Karlsruhe Institute of Technology, and Ruhr University Bochum. We aim at developing modern high-order finite element algorithms that can utilize the computational power of modern CPU and GPU hardware, which can be applied to challenging application problems in fluid dynamics.
Project partners
| PI | Research focus |
|---|---|
| Prof. Dr. Hartwig Anzt Karlsruhe Institute of Technology |
Linear algebra algorithms for GPUs Schwarz-type preconditioners with batched local solvers |
| Prof. Dr. Katharina Kormann Ruhr University Bochum |
Mixed-Precision Solvers for Implicit Time Stepping Structure-preserving Finite Element Methods |
| Prof. Dr. Martin Kronbichler University of Augsburg |
Node-level performance tuning of matrix-free algorithms Multigrid solvers for CPU and GPU architectures |
|
Technical University of Munich |
Cross-platform abstraction Large-scale scalability |
| Prof. Dr. Wolfgang A. Wall Technical University of Munich |
Development of CFD application solver ExaDG Application to LES to gas emission and biomedicine |
Open Positions
- PostDoc or Ph.D. position at University of Augsburg: Development of matrix-free algorithms for CPUs and GPUs, performance tuning, see the official announcement
- PostDoc or Ph.D. position at Ruhr University Bochum: Development of mixed-precision algorithms for time-dependent problems with structure-preserving finite elements, see the official announcement
-
PostDoc or Ph.D. position at Technical University of Munich (Department of Computer Engineering): Performance Evaluation and development of novel parallel programming abstractions for Exascale-class systems.