Practical parallel algorithms with MPI (PAMPI)5 ECTS
(englische Bezeichnung: Practical parallel algorithms with MPI)

Lehrveranstaltungen:

• • Practical parallel algorithms with MPI
    (Vorlesung, 2 SWS, Gerhard Wellein et al., Zeit und Raum n.V.)
  • Excercises for Practical parallel algorithms with MPI
    (Übung, 2 SWS, Gerhard Wellein et al., Zeit und Raum n.V.)

Empfohlene Voraussetzungen:

• Experience in C programming and basic knowledge of Linux CLI required.

Inhalt:

• Introduction to parallel programming and HPC system architecture
  

• MPI Basics: Point to point communication, Derived datatypes, collective communication operations
  

• MPI Advanced topics: Process topologies and MPI IO
  

• MPI parallel software design best practices
  

• Tools and debugging for MPI parallel programming
  

• Benchmarking of MPI runtime environments and applications
  

• Parallelisation of PDE solvers on structured grids on the example of a CFD solver
  

• Parallelisation of particle methods on the example of classical molecular dynamics

Lernziele und Kompetenzen:

Students

• acquire a comprehensive knowledge of parallel programming for distributed memory supercomputers
  

• learn the MPI library interface for point to point and collective communication and derived data types
  

• learn about advanced MPI topics, specifically process topologies and MPI IO
  

• learn how to benchmark, debug and profile MPI applications
  

• parallelize two non-trivial scientific computing applications: A CFD solver on structured grids and a classical molecular dynamics solver