Exploring load balancing in a scientific SPMD parallel application

We report the development of an SPMD parallel application which computes the macroscopic thermal dispersion in porous media. The performance of SPMD programs is strongly affected by dynamic load imbalancing factors. The use of a suitable load balancing algorithm is essential for overcoming the effects of these imbalancing factors. We developed nine versions of the SPMD application, each one adopting a different load balancing strategy. The main contribution of this work is the performance evaluation and comparison of these nine versions. The experimental results showed the importance of using an appropriate load balancing strategy for the characteristics of this scientific parallel application.