Siegel der Universität
Universität zu Köln

Mathematisch-Naturwissenschaftliche Fakultät
Mathematisches Institut

New Algorithms for Exascale Computing

Workshop at University of Cologne

December 04 - 06, 2013, Cologne, Germany

This workshop will address algorithmic, programming and application software aspects for the simulation of complex multiscale problems on the future supercomputers for exascale computing. It is organized within and funded by the German Priority Program 1648 "Software for Exascale Computing" (SPPEXA) by the Deutsche Forschungsgemeinschaft (DFG). Multiscale problems appear in different SPPEXA projects, e.g., the simulation of high strength steels in EXASTEEL or the simulation of flow in environmental or technical porous media in EXA-DUNE. ESSEX develops exascale sparse eigensolvers for studies of graphene-based structures, topological insulators, and quantum Hall effect devices. There is a common need for efficient computational homogenization schemes and highly scalable and robust preconditioners on the basis of algebraic multigrid and domain decomposition methods. In this workshop we plan to bring together internationally known experts from these fields of research with researchers from the projects ESSEX, EXA-DUNE, EXASTEEL, ExaStencils, and TerraNeo. The workshop will be also open for other projects from SPPEXA as well as for researchers from outside the priority program.

Photo of participants


Program and Schedule

Book of Abstracts

Invited Speakers

Slides (Password required)

Jö Fahlke, Performance of Thread-Parallel Assembly in Dune

Gundolf Haase, Accelerator programming and demands from applications

Michael A. Heroux, Toward the Next Generation of Parallel and Resilient Algorithms & Applications

Kab Seok Kang, A Fast parallel multigrid solver

David Keyes, Algorithmic Adaptations to Extreme Scale

Ulrike Meier Yang, Algebraic Multigrid – Getting Ready for Exascale

Steffen Müthing, Integrating multi-threading and accelerators into DUNE-ISTL

Robert Scheichl, Petascale Multigrid Performance and Beyond with Applications in the Earth Sciences

Jonas Thies, A Hybrid Parallel Block Jacobi-Davidson Method

Wim Vanroose, Hiding latencies in deflated Krylov solvers

Albert-Jan Yzelman, Parallel SpMV multiplication

Posters (Password required)

Albert-Jan Yzelman, Towards a highly scalable Multi-BSP fast Fourier transform

Organizing Committee