Oceananigans.jl: Improving Climate Model Accuracy with Fast and Friendly Geophysical Fluid Dynamics on GPUs
Event Type
ACM Student Research Competition: Graduate Poster
ACM Student Research Competition: Undergraduate Poster
Posters
Student Program
TP
XO
TimeThursday, 19 November 20208:30am - 5pm EDT
LocationPoster Module
DescriptionClimate models cannot resolve every cloud in the atmosphere and every wave in the ocean, but their collective effects can be important for global climate. Climate models rely on ad-hoc surrogate models to resolve the effects of these sub-grid scale physical processes. Many existing surrogate models are structurally deficient so they do not always respect physics and are the largest source of uncertainty in climate projections.
To aid in developing and training new, more accurate surrogate models, we developed Oceananigans.jl as a fast and friendly software package for the numerical simulation of incompressible, stratified, rotating fluid flows on CPUs and GPUs with large eddy simulation turbulence closures.
At SC20 I will highlight the climate research Oceananigans.jl is making possible and discuss performance benchmarks (multithreading and distributed CUDA-aware MPI). I will present some example scripts and show how Oceananigans.jl can be run on a laptop or cluster with the same script.
To aid in developing and training new, more accurate surrogate models, we developed Oceananigans.jl as a fast and friendly software package for the numerical simulation of incompressible, stratified, rotating fluid flows on CPUs and GPUs with large eddy simulation turbulence closures.
At SC20 I will highlight the climate research Oceananigans.jl is making possible and discuss performance benchmarks (multithreading and distributed CUDA-aware MPI). I will present some example scripts and show how Oceananigans.jl can be run on a laptop or cluster with the same script.