Recently the NASA Ames Research Center'9s NAS Systems Division and the Goddard Space Flight Center's Data Assimilation Office began a joint effort to dramatically increase the performance of NASA's next generation climate predicting Data Assimilation System (DAS). The effort is focused on inserting the highly scalable Multi- Level Parallelism (MLP) technique into the DAO codes.

The DAO office is responsible for the development of the science and the initial encoding of the algorithms. The NAS Systems Division is responsible for the adaptation of the codes to the SGI Origin platforms. The goal is provide a climate modeling system with performance substantially better than any other site currently in existence.

One of the major modules in the system is FVCORE, which computes the core dynamics of the NASA climate model, FVCCM3. Figures A and B on the left show the performance of this code at two grid resolutions on the NAS Systems Division's 512 CPU Origin 2000 system with contrasts to executions on other major compute platforms using MPI.