EXPERIENCES IN OPTIMIZING A NUMERCIAL WEATHER PREDICTION MODEL: AN EXERCISE IN FUTILITY?

This paper describes the basic optimization methods applied to a Numerical Weather Prediction model, including improved cache memory usage and message hiding. The goal is to reduce the wall clock time for time critical weather forecasting and compute intensive research. The problem is put into perspective via a brief history of the type of computing hardware used for NWP applications during the past 25 years and the trend in computing hardware that has brought us to the current state of code efficiency. A detailed performance analysis of an NWP code identifies the most likely parts of the numerical solver for optimization and the most promising methods for improving single and parallel processor performance. Results from performance-enhancing strategies are presented and help define the envelope of potential performance improvement using standard commodity based cluster technology. 1