csf-gpu-teaser.jpg

About this project

This project encompasses several efforts to port our community models to run on heterogeneous architectures, such as GPUs. These projects include the NSF CSSI-funded StormSPEED and TURBO projects, as well as the CSF-sponsored projects that port the MPAS-A and MUSICA models to run on GPUs. These projects use a variety of porting methods, including OpenACC (MPAS-A), CUDA (MUSICA), Kokkos (StormSPEED), and AMReX (TURBO).

Why this work is important

This work will make our community models portable across different HPC systems and enable us to run ultra-high-resolution simulations more efficiently.

How does this fit within the CSF

  1. Provides performance portability across different architectures.
  2. Allows us to run ultra-high resolution simulations more efficiently.

NSF NCAR Labs involved in this project

CISL
MMM
ACOM
CGD
 

External partners

University of Michigan
Texas A&M
NOAA GFDL
DOE

More information

StormSPEED
https://sites.google.com/umich.edu/nsf-stormspeed
https://github.com/NCAR/StormSPEED

TURBO
https://github.com/TURBO-ESM

MUSICA
https://www2.acom.ucar.edu/sections/multi-scale-infrastructure-chemistry-modeling-musica
https://github.com/NCAR/musica

MPAS-A
https://www2.mmm.ucar.edu/projects/mpas/site/index.html
https://github.com/MPAS-Dev/MPAS-Model