Project summary
The technologically driven change from ever faster single processors to multi-core processors with moderate single processor performance is taking place in full width: from the desktop to the supercomputer. For time-critical numerical simulations in particular, this means that computing power can only be achieved in the long term through new numerical methods or consistent optimization and massive parallelization. However, “tuning” the programs for high parallel computing power requires special knowledge that only very few research groups can build up by themselves and maintain in the long term. The present KONWIHR project OMI4papps (optimization, modeling, and implementation for parallel applications) addresses exactly this problem area by providing expert know-how in the optimization and parallelization of programs at a central location for all Bavarian HPC research groups. This includes in particular the support of other KONWIHR projects as well as users of the SuperMUC-NG at LRZ Garching. A training offer with courses on parallelization and program optimization complements these activities.
The HPC groups in Erlangen and Garching have long-term and proven expertise in the parallelization and optimization of application codes on all common (massively) parallel computers. In addition, there are close partnerships with industry partners such as Intel, which guarantee constant access to the latest computer architectures and software components relevant for high-performance computing.
KONWIHR funding
- OMI4papps is a follow-up project of RRZE’s cxHPC
- OMI4papps is also a follow-up project of LRZ’s previous support project
- KONWIHR funding of OMI4papps: since 9/2008
Contact:
- Dr. Matthias Brehm, LRZ Garching
- Prof. Dr. Gerhard Wellein, RRZE, FAU Erlangen-Nürnberg
Project staff:
- Ayesha Afzal, M. Sc., RRZE, FAU Erlangen-Nürnberg
- Dr. Ferdinand Jamitzky, LRZ Garching
Supported lectures
- Parallel Programming of High Performance Systems, yearly course, alternating between LRZ Garching and RRZE
-
PRACE Course: Advanced Topics in High Performance Computing, yearly course at LRZ Garching
Supported KONWIHR projects
- PEigFex: Highly parallel calculation of many internal and closely adjacent eigenvalues of fermionic and highly excited excitonic systems
- Integration of efficient compute kernels for phase field models into the waLBerla simulation framework using code generation and performance engineering techniques
- Efficient simulation experiments for parameter optimization of memory-intensive computer-linguistic learning methods
- Performance optimization of an iterative Maxwell solver
- Optimization and implementation of novel quantum Monte Carlo methods for strongly correlated electron systems
- Optimizing the Parallel Granular Gas Solver to study the crater formation
- Dynamics of Complex Fluids
Selected publications
- A. Afzal, G. Hager, and G. Wellein: Propagation and Decay of Injected One-Off Delays on Clusters: A Case Study. Accepted for IEEE Cluster 2019, Albuquerque, NM, September 23-26, 2019. Preprint: arXiv:1905.10603
- A. Afzal, G. Hager, and G. Wellein: Delay Flow Mechanisms on Clusters. Poster at EuroMPI 2019, Zurich, Switzerland, September 11-13, 2019.
- J. Hornich, G. Hager, and C. Pflaum: Efficient optical simulation of nano structures in thin-film solar cells. Proc. SPIE 10694, Computational Optics II, 106940R (28 May 2018); DOI: 10.1117/12.2312545
- C. Scheit, G. Hager, J. Treibig, S. Becker, and G. Wellein: Optimization of FASTEST-3D for Modern Multicore Systems. Technical Report (2013). Preprint: arXiv:1303.4538
- Further publications are available on the website of the HPC group.