Experience on the Parallelization of a Cloud Modeling Code Using Computer Aided Tools

The purpose of the work reported on is two-fold: The optimization of large-scale earth science applications and the evaluation of the use of automatic parallelization tools. We have applied the CAPO computer aided parallelization tool developed at the NASA Ames Research Center to the Goddard Cumulus Ensemble (GCE) cloud modeling code. We describe how the tool was used for OpenMP parallelization and identification of performance obstacles within the code. We also give performance results for a number of different hardware platforms, discuss the results, and outline possibilities for further optimizations in the code and improvement in the tools.