Performance Characteristics of Alogrithms from a Coupled Ocean/Atmosphere Model on Scalable Parallel Architectures.
A. Daniel Kowalski, Paul S. Schopf, Max J. Suárez · PPSC · 1993
Climate modeling, if done properly, requires an expenditure of an enormous amount of computing resources in both cpu time and mass storage. This is necessary, not only because of the need to adequately represent the fluid dynamics over a wide range of length and time scale, but also to include, as much as is possible, the important physical processes which influence the global climate. Thus, if the global ocean circulation, and the effects of sea ice and land surface details are represented in the model, as well as the atmospheric circulation, the combined coupled model can easily challenge the capacity of projected teraflop performance computers. The following describes some preliminary results from a project which will develop and evaluate algorithms on high performance architectures within the context of the coupled global climate modeling problem.