Paralell implementation of cellular fluid dynamics model

Davor Sluga (2010) Paralell implementation of cellular fluid dynamics model. EngD thesis.

PDF Download (1395Kb)

Abstract

In this thesis we developed a parallel implementation of cellular fluid dynamics model. Its advantage lies in very efficient exploitation of parallel computers. Firstly we took the equations which govern the fluid dynamics and developed a sequential algorithm. We then optimized the algorithm and improve its performance. We described different types of parallel computers and presented a methodology for developing parallel algorithms. We also defined metrics for assessing the performance of parallel algorithms. High efficiency was achieved by using the Monte Carlo method, which enables virtually linear speedup with increasing number of processors. Experiments which were performed on LSC Adria Supercomputer confirmed this prediction. We observed how changing the algorithms parameters affects the quality of the result and processing time. The results were consistent with theoretical predictions consequently parallel algorithm proved to be very effective.

Actions (login required)