Set of MITgcm runs
-
Installation prerequisites on any debian-like system (ubuntu, etc.) sudo apt-get install gfortran sudo apt-get install libnetcdff-dev libnetcdf-dev sudo apt-get install python-numpy python-matplotlib
-
Dowload MITgcm in a local directory
git clone https://github.com/MITgcm/MITgcm.git
cd MITgcm
echo " " >> ~/.bashrc # empty line (safety)
echo "export MITGCM_ROOTDIR=$PWD" >> ~/.bashrc
echo "export PATH=\$PATH:\$MITGCM_ROOTDIR/tools" >> ~/.bashrc
echo "export PATH=\$PATH:\$MITGCM_ROOTDIR/utils/python/MITgcmutils/scripts" >> ~/.bashrc
cd ..
source ~/.bashrc
- In a separate repository get mitgcm_configs
git clone https://github.com/bderembl/mitgcm_configs
cd mitgcm_configs
- In each configuration (e.g. for 'corner' here)
cd corner
mkdir build
mkdir run
- Build mitgcm
At this point you need a fortran compiler (and openmpi for big runs) In most HPC configurations, you will need to load the appropriate module. Check
module avail
and load the compiler you want (something like):
module load gnu
module load openmpi
cd build
genmake2 -mods=../code (-mpi)
make depend
make -j4
(Use the '-mpi' option for parallel runs. If genmaka2 returns an erros, you may have to adjust set the variable MPI_HOME in your .bashrc; use the command which mpif77 to check where this folder is)
- Create init files
cd ../input
python mygendata.py
- Run the model
cd ../run
ln -s ../input/* .
cp ../build/mitgcmuv .
./mitgcmuv