Research Centre JuelichCentral Institute for Applied Mathematics : Research Centre Juelich Central Institute for Applied Mathematics
Central Institute for Applied Mathematics : Central Institute for Applied Mathematics Planning and operation of the central computing systems for the research center (FZJ) and the John von Neumann-Institute for Computing (NIC)
Planning and operation of the data communication systems within FZJ
Research and development in mathematics, data processing, and computer science
Consulting as well as training in mathematics, data processing, and networking
Education of Mathematical-technical Assistants
John von Neumann-Institute for Computing (NIC) : John von Neumann-Institute for Computing (NIC) founded in 1998
Partners: DESY (Hamburg), FZJ (Jülich)
Tasks:
provide supercomputer power to german research groups
evaluate new computer designs
increase the collaboration of computer scientists and application experts
Research Background : Research Background performance analysis tools: VAMPIR, OPAL, PAT
performance analysis meta-tools: EARL
programming models: SVM-Fortran
applications: NIC
parallel architectures: SM, DSM, MP
KOJAK : KOJAK program information
trace data
summary information
scalability information Database scalabity analysis summarizer prediction tool compiler instrumenter source analyzer trace monitor profiling tool Auto Instrumenter Auto Executor Auto Finder Auto Controller proven hypotheses hypotheses bottlenecks KOJAK user-supplied
hypotheses application
The Experts : The Experts Michael Gerndt
Forschungszentrum Jülich GmbH Zentralinstitut für Angewandte Mathematik D-52425 Jülich Phone: +49 (2461) 61 6569 Fax: +49 (2461) 61 6656 email: m.gerndt@fz-juelich.de
Bernd Mohr
Forschungszentrum Jülich GmbH Zentralinstitut für Angewandte Mathematik D-52425 Jülich Phone: +49 (2461) 61 2054 Fax: +49 (2461) 61 6656 email: b.mohr@fz-juelich.de
Planned Main Contributions to APART : Planned Main Contributions to APART identification and description of bottlenecks (MP, DSM)
implementation designs for future environments
Expected Achievements : Expected Achievements collaboration with excellent groups
information to be used for the design of KOJAK
selection of a good representation technique for bottlenecks
catalog of bottlenecks
design partterns of generic automatic environments