New DFG Projects GROMEX and CATWALK
Within the DFG priority programme 1648 "Software for Exascale Computing" (SPPEXA), JSC is partner in two new projects GROMEX and CATWALK. Both projects are being funded for a period of three years.
The goal of the GROMEX project is to develop a flexible, portable and ultra-scalable solver for potentials and forces within the GROMACS MD code, which is a prerequisite for exascale applications in particle-based simulations with long-range interactions in general. The second challenge is to realistically describe the time-dependent location of (partial) charges, as e.g. the protonation states of the molecules depend on their time-dependent electrostatic environment. Both tightly interlinked challenges are addressed by the implementation and optimization of a unified algorithm for long-range interactions that will account for realistic, dynamic protonation states, and at the same time overcome current scaling limitations.
The project is coordinated by the Max Planck Institute for Biophysical Chemistry in Göttingen with JSC and the Stockholm University as project partners. JSC will provide an implementation of a linear scaling, error-controlled fast multipole method (FMM), which will be ported to the BlueGene/Q architecture and methodically extended to allow for dynamic protonation states without introducing additional computational overhead. For more information, visit the website http://www.mpibpc.mpg.de/grubmueller/sppexa or contact Dr. Holger Dachsel (h.dachsel@fz-juelich.de).
The CATWALK project deals with the following aspects: The cost of running applications at exascale will be tremendous. Reducing the runtime and energy consumption of a code to a minimum is therefore crucial. Moreover, many existing applications suffer from inherent scalability limitations that will prevent them from running at exascale in the first place. Current tuning practices, which rely on diagnostic experiments, have drawbacks because they detect scalability problems relatively late in the development process when major effort has already been invested in finding an inadequate solution. Furthermore, they incur the extra cost of potentially numerous full-scale experiments. Analytical performance models, in contrast, allow application developers to address performance issues already during the design or prototyping phase. Unfortunately, the difficulties of creating such models combined with the lack of appropriate tool support still render performance modelling an esoteric discipline mastered only by a relatively small community of experts.
The project's objective is, therefore, to provide a flexible set of tools to support key activities of the performance modelling process, making this powerful methodology accessible to a wider audience of HPC application developers. The partners are the GRS Aachen, Technische Universität Darmstadt, Goethe Center for Scientific Computing, Swiss Federal Institute of Technology Zurich, and JSC. More information can be found at http://www.vi-hps.org/projects/catwalk or contact Dr. Bernd Mohr (b.mohr@fz-juelich.de).
from JSC News No. 211, 15 March 2013