The aim of this workshop is to give people an insight into the programming of parallel machines by MPI. The workshop is split into two parts. During the first part (three half-days) an introduction to MPI-1 will be given. Hands-on sessions will allow users to immediately test and understand what the basic constructs of MPI are meant to do. The second part (two half-days) will be dedicated to more sophisticated techniques of programming in MPI-1 and MPI-2.

This 2½ day course is given four times at three locations:

• Jülich, John von Neumann Institute for Computing (NIC)
  • 2000, Wednesday March 29, 13:00 - Friday March 31, 16:30,
• Stuttgart, High-Performance Computing-Center Stuttgart (HLRS),
  • 2000, Monday April 3, 9:00 - Wednesday April 5, 12:30,
  • 2000, Wednesday April 5, 13:00 - Friday April 7, 16:30,
• Dresden, Center for High Performance Computing (ZHR).
  • 2000, Monday April 10, 9:00 - Wednesday April 12, 16:30 (the Tuesday morning is free to have a look at the famous city of Dresden)

The course language is German (except the course April 5-7 at HLRS, which is in English). All slides and course materials are in English.

Rolf Rabenseifner
Phone: 0711 685 5530
E-mail: rabenseifner@hlrs.de

Local organizer in Jülich: Norbert Attig, phone 02461 61 4416, sc.zam@fz-juelich.de
Local organizer in Dresden: Claudia Schmidt, phone 0351 463 4274, schmidt@zhr.tu-dresden.de

1. half-day-block: Parallel Programming
   • Introduction (Talk)
   • Parallel Programming Models (Talk)
   • MPI - Getting started (Talk)
   • MPI Programs (Talk and Practical Session)
   • What's in a message? (Talk)

2. half-day-block: MPI Message Passing
   • Point-to-Point Communication (Talk and Practical Session)
   • Non-Blocking Communication (Talk and Practical Session)
   • Derived Datatypes (Talk and Practical Session)

3. half-day-block: MPI collective and more sophisticated constructs
   • Virtual Topologies (Talk and Practical Session)
   • Collective Communication (Talk and Practical Session)
   • Other MPI-1 features and MPI-2 overview

4. half-day-block: MPI File I/O and optimization hints
   • Parallel File I/O with MPI-2 (Talk and Practical Session)
   • Optimization of MPI Applications
     • Overview and Introduction (Talk)
     • MPI Performance Studies (Talk)

5. half-day-block: Application example and profiling tools
   • Optimization of MPI Applications (continued)
     • a CFD example (Talk and Practical Session)
     • Counter-based Profiling (Talk)
     • Trace-based VAMPIR Profiling (Talk and Practical Session)
   • Access to the federal high-performance computing-centers

The course materials are handed out to all participants. You can get additional copies from the course chair and editor Rolf Rabenseifner. Here you can get all course material on the WEB. The MPI-1 part is based on the MPI course developed by the EPCC Training and Education Centre, Edinburgh Parallel Computing Centre.