Title:

Background:

Large scale numerical simulations make great demands on computing power and available memory. Parallelization on workstation clusters is one answer to these increasing requirements and the Message Passing Interface (MPI) has become the standard for distributed memory parallelization. IMMD10 is carrying out research in this area with its cluster of XP1000 (alpha) workstations.

Typical examples of large scale problems come from computational fluid dynamics (CFD). CFD allows to simulate and predict the aerodynamic or hydrodynamic properties in many technical applications where the transport in or of fluids such as air and water is of importance. Notable examples are the aerodynamics of aircraft, passenger cars or turbomachinery. Other important fields of interest are related to applications in thermodynamics, combustion and chemistry.

In the summer term 1999 the chairs for System Simulation (IMMD10) and for Fluid Mechanics (LSTM) organized a course on Numerical Simulation of Fluids, during which a comparatively simple, but very flexible solver for the Navier-Stokes equations, the governing equations of CFD, was developed.

Task:

The task of this Studienarbeit / Master Thesis is

• Parallelization of the CFD code (which is written in C) using the MPICH library (a free MPI implementation).
• Testing of the parallel code on the workstation cluster of IMMD10.
• Performance analysis of the code for different communication interfaces (Fast Ethernet versus Myrinet).

Contact:

Christoph Freundl is working on this project, his advisor is Marcus Mohr.