Climate Modelling Lab, MSc Meteorology S2025 (280351 VU)

Recipe for lecturer: generate a tarball with the icon model code for distribution to students

Work with VSC account avoigt_teach.

Get model code, use branch climmodlab_s2023:

git clone https://gitlab.phaidra.org/climate/icon-esm-univie.git
git checkout origin/climmodlab_s2023

Configure following the recipe for the students below. Then create a distribution tarball:

make dist

This leads to the file icon-2.6.0.tar.bz2, which can be distributed to the students.

VSC reservation info and usage

To check the reservation:

scontrol show reservation=p72044

To use the reservation on the command line:

--reservation=p72044

To use the reservation in a job script:
#SBATCH --reservation=p72044

Recipe for students: how to get, configure and compile ICON on VSC4

Reserve a node for interactive use, do not work on the login nodes. See VSC Wiki.

salloc -N 1 --qos=skylake_0096 --partition=skylake_0096 --reservation=p72044 --exclusive -t 2:00:00

The ssh into the reserved node. You can check the reserved node with

squeue -u $USER

Get and untar the tarball and enter the model code directory:

cp  /gpfs/data/fs72044/avoigt_teach/icon-2.6.0.tar.bz2 /gpfs/data/fs72044/iconXY
tar xfv icon-2.6.0.tar.bz2
mv icon-2.6.0 icon-esm-univie
cd icon-esm-univie/

Configure the model, using MPI parallel communication (this step typically takes several minutes):

config/vsc/vsc4.intel-intelmpi_spack --enable-mpi

You can check for a successful configure step by looking at the last line of the file config.log. This line should read configure: exit 0.

Load the libraries that are needed for compilation of the model:

spack load /xg435ds  # intel-oneapi-compilers@2022.2.1 linux-almalinux8-skylake_avx512 / gcc@9.5.0
spack load /p7jisxw  # intel-oneapi-mkl@2022.2.1 linux-almalinux8-skylake_avx512 / intel@2021.7.1
spack load /6a7vghr  # netcdf-fortran@4.6.0 linux-almalinux8-skylake_avx512 / intel@2021.7.1
spack load /kqslczm  # libxml2@2.10.1 linux-almalinux8-skylake_avx512 / intel@2021.7.1
spack load /tlvfj62  # eccodes@2.25.0 linux-almalinux8-skylake_avx512 / intel@2021.7.1

Compile the model, using 8 cores for faster compilation (this step typically takes several minutes):

make -j 8

A successful compilation should end with the following lines printed to screen

FCLD bin/icon

support/util_file.o: In function `util_tmpnam':

/home/fs72044/avoigt_teach/climlab_s2024/icon-esm-univie/support/util_file.c:63: warning: the use of tmpnam' is dangerous, better use mkstemp'

make[1]: Leaving directory '/home/fs72044/avoigt_teach/climlab_s2024/icon-esm-univie'

The compilation leads to the binary bin/icon.

Running an AMIP-like simulation

A runscript for an AMIP-like simulation is available in runs/amip/, the run script is exp.amip.run. The script is written for user avoigt_teach and needs to be adapted accordingly for others users (e.g., replacing avoigt_teach where needed and adapting some of the directory names.)

The simulation is AMIP-like because SSTs, sea ice etc. are prescribed as climatological averages and are the same for each year.