Achieving Performance for Paleoclimate Models
Climate modelers use a hierarchy of models, ranging from box models to 3D, to address a huge range of time and space scale, from cloud microphysics to ice-sheet dynamics, and from seconds to millennium, or even more. This implies using the largest computers available, or places a heavy burden on the parameterization of unresolved processes.
In paleoclimate studies, the length of the simulations (century to multi-millenia) restrict 3D models to coarse resolution and small size of grids, constraining parallelization possibilities. In this talk, I will present the scientific goals, and the strategy to achieve computing performance.
Olivier Marti, Institut Pierre Simon Laplace, France