FloeDynHub / FloeDyn

FloeDyn codebase
Apache License 2.0
3 stars 2 forks source link

Todo: a few words about floedyn ...

For more information, please visit the FloeDyn documentation.

Download and install

Prerequisites

Get sources

> git clone git@github.com:FloeDynHub/FloeDyn.git

Configure

Assuming that is the path to the Floe_Cpp repository Run:

> cd <source_dir>
> python3 ./waf configure --prefix=<where-you-want-to-install> 

Results:

By default, waf looks for dependencies in /usr/. To specify another location (e.g. on clusters using guix or nix), try

> python3 ./waf configure --prefix=<where-you-want-to-install> --default-search-path=<other-path>
# eg --default-search-path=$GUIX_PROFILE on Luke or Dahu

More about waf: https://waf.io/apidocs/tutorial.html

Build

> python3 ./waf --target FLOE -v
# -v is verbose mode, optional
# use --target FLOE_MPI for the parallel version

Execute

mpirun -np 2 <path-to>/build/FLOE_MPI -i <input.h5> -t <nb_time_steps>

Some h5 files are available in Floe_Cpp/io/inputs/.

Build and install on Gricad clusters

Install all dependencies using guix inside a 'floedyn' profile:

source /applis/site/guix-start.sh
guix package  -m ./Floe_Cpp/manifest_floedyn.scm -p $GUIX_USER_PROFILE_DIR/floedyn

Or work inside a guix environment

source /applis/site/guix-start.sh
guix environment  -m ./Floe_Cpp/manifest_floedyn.scm --pure

Configure and build floedyn

In both case, use the same process as described above (waf configure and so on) with the --default-search-path=$GUIX_PROFILE option.

Execute

An example of a oar script is given in run_simu.sh.

Try

oarsub -S "./run_simu.sh $HOME/Floe_Cpp/io/inputs/in_single_small_floe.h5 20"