Skip to content

Using AI compilers to optimize exascale climate simulations on traditional and cloud HPC systems

Notifications You must be signed in to change notification settings

PRONTOLab/GB-25

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

GB-25

This repository accompanies a submission for the 2025 Gordon Bell climate prize submission that showcases Reactant-acceleration of Oceananigans, ClimaOcean, and SpeedyWeather simulations.

Package organization

  • src implements two models:

    • GordonBell25.data_free_climate_simulation_init, which uses ClimaOcean to drive an Oceananigans HydrostaticFreeSurfaceModel
    • GordonBell25.baroclinic_instability_model, a simpler, unforced, pure Oceananigans setup.
    • Both models can be to use either a LatitudeLongitudeGrid or a TripolarGrid with idealized bathymetry.
  • simulations: scripts to i) compile and ii) run either of the two models

  • sharding: scripts and utilities that use XLA's sharding to distribute computations across multiple nodes. Oceananigans uses its Distributed architecture to represent sharding across nodes.

  • ext: many small packages that each precompile part of a model time-step, in order to accelerate compilation during intensive jobs.

About

Using AI compilers to optimize exascale climate simulations on traditional and cloud HPC systems

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages