-
Notifications
You must be signed in to change notification settings - Fork 78
HOWTO install on Genius (for KU Leuven users)
If you are going to use a GPU-enabled version of COOLFluiD on Genius in the VSC cluster with your KU Leuven account, you don't need to install all the dependencies since they are already installed. That means that you DO NOT need to follow all the steps explained here. Instead, you want to follow the steps below:
- 1.- Make sure you have an VSC/Genius account and you can access it from your terminal. (Check this link if you don't have an account yet)
-
2.- Run
wget https://raw.githubusercontent.com/andrealani/COOLFluiD/master/tools/conf/Genius/install_COOLFluiD.sh
-
3.- (OPTIONAL) Edit the file
install_COOLFluiD.sh
and set the variables:
TOP_DIR
: your data directory where COOLFluiD is downloaded (default: ${VSC_DATA}
);
COOLFLUID_TOP_DIR
: the folder where COOLFluiD source code has been downloaded from github (default: ${TOP_DIR}/COOLFluiD_Genius
).
-
4.- Make the installation script runnable:
chmod u+x install_COOLFluiD.sh
- 5.- Run:
./install_COOLFluiD.sh DEBUG_CUDA --download=2
(with debugging, development version, with CUDA)
or
./install_COOLFluiD.sh OPTIM_CUDA --download=2
(w/o debugging, production runs, with CUDA)
or
./install_COOLFluiD.sh DEBUG_NOCUDA --download=2
(with debugging, development version, w/o CUDA)
or
./install_COOLFluiD.sh OPTIM_NOCUDA --download=2
(w/o debugging, production runs, w/o CUDA)
to download, configure, compile and install COOLFluiD.
NOTE: If you don't want to download COOLFluiD but just recompile it, use --download=0
(only update, no configuration via prepare.pl
) or --download=1
(update and configuration via prepare.pl
) instead of --download=2
in the previous command.
Normally, the installation should work out-of-the-box without step (3), since the default settings are good enough.
NOTE: Sometimes the downloading of COOLFluiD is slower than usual (connection issues?) and timed out on VSC. In this case only part of the source code is downloaded, and often the .conf
file needed for the installation is not there yet, giving you an error when trying to configure/compile. In this case, please do:
svn cleanup COOLFluiD_Genius ; ./install_COOLFluiD.sh DEBUG_CUDA --download=0
(in here use the same *CUDA as before)
as many times as needed for your COOLFluiD download to be fully successful (it can be many times...). Then, in order to complete the installation run:
./install_COOLFluiD.sh DEBUG_CUDA --download=1
(in here use the same *CUDA as before)
If you experience problems, before contacting VSC system support, please contact andrea.lani@gmail.com or andrea.lani@kuleuven.be.
In order to run on a cluster node in production (i.e. fully optimized) mode, you will have to load the following modules, either inside your PBS script or your .bashrc
:
module load CMake/3.10.2-GCCcore-6.4.0
module load paralution/1.1.0-foss-2018a
module load Boost/1.66.0-foss-2018a
module load ParMETIS/4.0.3-foss-2018a
module load PETSc/3.9.0-foss-2018a-CUDA
(with CUDA)
or
module load PETSc/3.9.0-foss-2018a-cpu
(w/o CUDA)
The COOLFluiD team is involved and open to new collaborations in Horizon2020, ESA and other research projects dealing with complex multi-disciplinary problems and computational challenges.
Training sessions and consulting services can also be provided on demand.
For any request or suggestions please contact andrea.lani@gmail.com
Parallel mesh decomposition
High-performance computing (strong scaling on NASA Pleiades for 1/2 billion-cells 3D grid)
Chemically reacting flows and plasma
Complex all-speed flow simulations