Parallel Computing in the Computer Science Curriculum > Modules > MPI Programming Exemplars

MPI Programming Exemplars

Hannah Sonsalla and Libby Shoop, Macalester College
Author Profile

Summary

Four examples of using MPI to build distributed programming solutions to classical problems:

  • Estimating Pi using a Monte Carlo method
  • Estimating an integral using the trapezoidal rule
  • Odd - Even transposition sort
  • Merge sort

Learning Goals

  • To study example problem solutions that use message passing.
  • To study example problems that employ the reduction pattern.
  • To see examples of how to time MPI programs and determine their speedup and efficiency compared to sequential solution.

Context for Use

Any one or all of these examples could be used after students have learned basic message passing patterns and have a brief familiarity with MPI programming. Students can run these examples and use a spreadsheet (linked below) to determine the speedup and efficiency of these programs for varying numbers of processes and problem sizes.

Description and Teaching Materials

The descriptions of the problems and links to the code are in this web page:

Message Passing with MPI

We suggest having students copy this Google spreadsheet and fill in the times the get with different problems sizes and numbers of processors.

As an alternative, you could use a paper version of this Google doc to have students fill in their times and draw the speedup and efficiency curves.

Teaching Notes and Tips

Any one of these works as an interesting example. The sorts ore more complex that the monte carlo and integration examples.


Comment? Start the discussion about MPI Programming Exemplars