The Leibniz Supercomputing Centre (LRZ) of the Bavarian Academy of Sciences and Humanities and the Department of Informatics at the Technische Universität München (TUM) along with Intel Corporation have jointly established an Intel Parallel Computing Center (IPCC) called “Extreme Scaling on MIC/x86.”
There are already a number of Intel Parallel Computing Centers existing and their mission is to facilitate software development of highly parallel programming codes for future supercomputers, especially those comprised of a very high number of processors with co-processors.
The trend in supercomputing over the last several years has been toward processors with an increasing number of cores which are combined to form high-performance compute nodes. These modern supercomputers are characterised by declining memory per core, especially those relying on wide vector units. In addition, increasing numbers of computing operations are being transferred to coprocessors such as Intel’s Xeon Phi coprocessor, which are able to compute specific problems much more quickly and more energy efficiently.
Developing and optimizing software for these highly parallel supercomputers with coprocessors allows for taking full advantage of the new technologies and achieving very significant increase in performance and energy efficiency. The IPCC founded by the LRZ, TUM, and Intel will tackle this challenge by optimizing four applications which are already running with excellent performance on SuperMUC (the supercomputer at the LRZ) for the next computer generation.
Within the coming months, SuperMUC will undergo its scheduled expansion: one part of this will be based on Intel Xeon Phi™ coprocessors . The respective combination of processors and co-processors strongly relies on efficient software parallelization and will take developments in parallel software into account in the coming years in order to reach even faster computing systems.
The four software packages to be optimized simulate earthquakes and seismic wave propagation (SeisSol), study the development of the cosmos (GADGET), apply molecular modeling methods to industrial applications (ls1 mardyn), and address high dimensional problems (SG++) that occur in data mining or in financial mathematics. All these programs are already running on supercomputers, specifically on the SuperMUC, with excellent performance in the Petaflops range. They now need to be optimized to run on computer systems in the range of hundreds of Petaflops and more.
The Intel Parallel Computing Center also aims beyond the further development of these simulation programs. Ultimately, its goal is to understand the process of developing scientific software for future computing systems, both in terms of optimizing the computational power as well as the energy efficiency, and to develop a model that captures this knowledge. The results and conclusions from this endeavor will merge into the ongoing development of the four simulation programs and will also be published in scientific journals for public dissemination.
“The expertise in the field of parallelization of applications at the TUM and the LRZ in conjunction with the Intel Parallel Computing Center Program is a combination with phenomenal synergies,” affirms Stephan Gillich, HPC Director EMEA, Intel GmbH “Systems with ever-increasing computational power which are based on parallel structures will help us to solve important challenges in scientific as well as industrial arenas.”
According to Professor Arndt Bode, Chairman of the Board of the LRZ, “With this Parallel Computing Center, the Leibniz Supercomputing Centre of the Bavarian Academy of Sciences and Humanities, the Informatics Department of the TUM, and Intel Corporation are working together to broaden their competence for the efficient operation of future supercomputers in nearly all scientific arenas. “
The TUM and the LRZ are represented in the Intel Parallel Computing Center “Extreme Scaling on MIC/x86” by Professors Michael Bader, Arndt Bode, and Hans-Joachim Bungartz.
Dr. Ludger Palm
Telephone +49 89 35831 8792
The Leibniz Supercomputing Centre (LRZ) of the Bavarian Academy of Sciences is the general IT service provider to the scientific and academic communities in the Munich area. LRZ operates the Munich Scientific Network (MWN), offers archiving and backup of large amounts of data and delivers high performance and supercomputing facilities for all German universities and PRACE partners via the Gauss Centre for Supercomputing (GCS).
Dr. Ellen Latzin | idw - Informationsdienst Wissenschaft
18.08.2017 | Albert-Ludwigs-Universität Freiburg im Breisgau
AI implications: Engineer's model lays groundwork for machine-learning device
18.08.2017 | Washington University in St. Louis
16.08.2017 | Event News
04.08.2017 | Event News
26.07.2017 | Event News
18.08.2017 | Life Sciences
18.08.2017 | Physics and Astronomy
18.08.2017 | Materials Sciences