Developing Supercomputing Geoscience Applications Using Julia

Explore recent efforts to accelerate geocomputing using Julia at scale in this conference talk from JuliaCon 2024. Learn about the modular approach developed through the PASC-funded GPU4GEO project, in collaboration with the Swiss National Supercomputing Centre CSCS. Discover key packages like GeophysicalModelGenerator.jl for initial model configuration, GeoParams.jl for material parameters and constitutive relationships, and JustPIC.jl for particle-in-cell advection on multi-GPUs. Understand the advantages of this composable framework in lowering barriers for new users to contribute to open-source packages. Examine application codes such as MagmaThermoKinematics.jl, JustRelax.jl, and FastIce.jl that utilize this functionality for various geodynamic simulations. See how these HPC building blocks enable geoscientists to write high-performance stencil codes scalable from laptops to large GPU-accelerated supercomputers. Gain insights into preliminary results on using automatic differentiation for inverse modeling and discuss how these developments will accelerate geocomputing and shape the education of future geoscientists.