Building a High-Performance Earth System Model in Julia

We are very pleased to have Maciej Waruszewski speak to us on the wonderful Climate Modeling Alliance (CliMA) [1][2] project where he is working on the use of Julia in solvers on both CPUs and GPUs. CliMA is a coalition of scientists, engineers, and applied mathematicians from Caltech, MIT, the Naval Postgraduate School, and NASA’s Jet Propulsion Laboratory. Maciej is a graduate of the Faculty of Physics, University of Warsaw, and is a postdoctoral research scholar at the Applied Mathematics department of the Naval Postgraduate School [3]. Maciej's abstract follows:

"To reduce and quantify uncertainties in climate predictions the Climate Modeling Alliance [1] aims to develop the first Earth system model that will automatically learn from observations and targeted high-resolution simulations. The new model is being developed in Julia to ensure performance on modern heterogeneous architectures without sacrificing scientific productivity. This talk will start with a quick overview of the project and then focus on its approach to performance portability. I will share our experiences with writing fast Julia code for both CPUs and GPUs and present some of the lessons learned."

We'll have pizza at 6:00, with the talk starting at 6:30.

CliMA's code is on Github [4]; see also a JuliaCon 2019 talk [5] on CliMA.

[1] https://clima.caltech.edu/
[2] https://www.sciencemag.org/news/2018/07/science-insurgents-plot-climate-model-driven-artificial-intelligence
[3] https://www.nps.edu/web/math/
[4] https://github.com/climate-machine/CLIMA
[5] https://www.youtube.com/watch?v=gD5U_U9kZk8