Global Storm Resolving Simulations in a Python Framework - Progress and Challenges
DescriptionThe modern exascale computing systems have given the climate and weather modeling community an opportunity for a step change in the simulation capabilities. One expects a resolution high enough to resolve the key atmospheric processes provided the machine is properly utilized. However, this is a non-trivial task. Given the complexities at various levels from the hardware to the software (model), a close collaboration between the computer scientists, software engineers, and domain experts is required to redesign the conventional model in a way that fully utilizes the exascale capabilities while keeping the model easy to use. In EXCALIM, we are developing a python-based framework that allows users to write code, debug, run the model, and visualize from within the framework. Computations are largely handled by the embedded DSL, GT4Py, whereas the driver code is envisioned to be in python. This approach decouples the hardware-specific implementation choices made for efficiency, from the functional choices by the domain experts. The framework is expected to allow for interoperability with software components developed in other programming languages or other frameworks. The talk will give an overview of the progress we have made in the last year and the challenges we have encountered.
TimeWednesday, June 2812:30 - 13:00 CEST
Climate, Weather and Earth Sciences