Page Comparison

...

1. Create a conda environment with mpas_tools and the no-MPI verison of ESMF:

   Code Block
   mamba create -y -n mpas_tools python=3.11 "mpas_toolstools>=0.20.0" "esmf=8.2.0=nompi*" mamba activate mpas_tools

   
   

2. Run the sea ice graph generation tools

   Download

   Code Block
   cd /lcrc/group/e3sm/public_html/inputfiles/ice/mpas-seaice/<mesh_name> simple_seaice_partitions -m <mesh_file> -p partitions/mpas-seaice.graph.info.<creation_date> -n <core_count> <core_count> <core_count>

   If you are not on LCRC, you need to download the required input files somewhere:
   https://web.lcrc.anl.gov/public/e3sm/mpas_standalonedata/mpas-seaice/partition/seaice_QU60km_polar.nc
   https://web.lcrc.anl.gov/public/e3sm/mpas_standalonedata/mpas-seaice/partition/icePresent_QU60km_polar.nc
   

   Run the sea ice graph generation tools

   Code Blockprepare_seaice_partitions -i seaice_QU60km_polar.nc -p icePresent_QU60km_polar.nc -m <mesh_file> -o . create_seaice_partitions -m <mesh_file> -o . -p mpas-seaice.graph.info.<creation_date> -n <core_count> <core_count> <core_count>

   and then point to them with the -d.

The mesh file is a ifle file with the mesh you are trying to partition, typically seaice.*.nc or mpassi.*.nc in inputfiles/ice/mpas-seaice/<mesh_name>.  The creation date needs to match the creation date of other sea ice partition files in the partitions directory (or it won't be found).  You can provide multiple core task counts in one call to createsimple_seaice_partitions, which will be much faster than calling that tool multiple times.  Put the resulting mpas-seaice.graph.info.<cration_date>.part.<core_count> files in /lcrc/group/e3sm/public_html/inputfiles/ice/mpas-seaice/<mesh_name>The tool should take care of setting permissions as well.