To better assess the impact of a warmer world, coupled climate models such as ICON must be capable of global simulations at the km scale. The move to higher resolution and thus more grid points is accompanied by a significant increase in the resources required. In both the compute and data analysis domains, adapted methods on HPC systems will be necessary to enable global, storm-resolution simulation. As the HPC market has evolved in recent years into a heterogeneous hardware environment with diverse software tools, existing codes are no longer necessarily executable on new systems. Similarly, the user community's demand for a climate model has grown, making it difficult to incorporate external components into the existing software design.
The primary goal of the WarmWorld-Faster project is to reengineer the infrastructure and key components of ICON to enable scalable development and application on exascale HPC systems. This involves meeting requirements for performance, flexibility and usability of the model, which requires a move away from monolithic design. Rather, a transformation to a modular, flexible model code is required. In this way, both individual components can be improved and adaptations to new programming paradigms can be made without having to adapt the entire model code.
