Contrib:Claws/code saturne

From CAELinuxWiki
Jump to: navigation, search
                     Installing Code_Saturne 1.4b on
                     (Ubuntu) GNU/Linux
                     September 18, 2008



This are some notes on installing Code_Saturne 1.4b on a Ubuntu 8.04 system. I'll try to make the procedure as generic as possible, so anyone with any Linux-system can use it. It provides a good basic Code_ Saturne install that supports most options. Code_ Saturne is highly configurable - but we won't go there in this document. This install-guide requires next to no experience to follow.

Other more detailed documents can be found or will be created in the near future.

It is created with the help of David Monfort of the Code_ Saturne development team and forum, who has been surprisingly patient with me. Always check the documentation before resorting to violence.

      • For ubuntu 8.10 you have to manually install Metis-4.0 (see how CGNS is installed) since metis on 8.10 is a wrapper for SCOTCH and wont work properly.

Obtaining Code_Saturne

Downloading Code_Saturne

Preparing the system.

Using the package manager on your system, you can install the optional/dependant libraries rather than doing a manual install.

Here's a list of needed libraries for this install - some might be installed by default on your system:

  • libmedc1
  • libmedc-dev
  • libhdf5-serial
  • libhdf5-serial-dev
  • libopenmpi1
  • libopenmpi-dev
  • libxml2
  • libxml2-dev
  • Zlib
  • Tcl/Tk/Tix
  • Python

Now, please note that libhdf5-serial is not the same as libhdf5-mpich/LAM.

This simple command in Ubuntu, installs the needed libraries:

claus@claus-laptop:~$ sudo apt-get install libopenmpi-dev
libmedc-dev libxml2-dev libparmetis-dev libhdf5-serial-dev
gfortran tix

This should take care of most business - cgns has to be installed manually:

tar zxf cgnslib_2.5-3.tar.gz && cd cgnslib_2.5/ && ./configure --prefix=/usr && make && sudo make install

Installing Code_Saturne 1.4b

I am installing Code_Saturne in /home/user/code_saturne, but it does not matter where you put it.

Creating the folders (assuming you are in home/user/ ):

mkdir code_saturne
mkdir code_saturne/Enveloppe - extract ecs-14a.tgz here
mkdir code_saturne/Interface - extract ics-14a.tgz here
mkdir code_saturne/Noyau - extract ncs-14b.tgz here, patch-ncs-14b_080909.tgz should be extracted here as well
mkdir code_saturne/src - bft-107.tgz and fvm-0110.tgz is extracted here

Next you should extract the install.tgz in code_saturne/ creating Install/

So far so good.

- go to the directory $INSTALLPATH/Noyau/ncs-x.y.z/bin
- open the cs_profile file and set the following variables:
CS_HOME -> home directory of the Kernel
ECS_HOME -> home directory of the Preprocessor
CSGUI_HOME -> home directory of the
GUI SYRCS_HOME -> home directory of the library used for coupling
with the SYRTHES solid thermal code
- source the cs_profile to set the environment variables
properly (usually type ". ./cs_profile") (in the
$INSTALLPATH/Noyau/ncs-x.y.z/bin folder)

Edit code_saturne/Install/setup with your favorite editor and change (the first one in the row) no to yes on the libraries you want to be used with Code_Saturne - metis, med, hdf, openmpi (for parallel processing) are good places to start.

Next edit code_saturne/Noyau/ncs-1.4b/bin/ (or if you are on a 64bit system etc.)to suit your needs - here you change the root of your code_saturne folder etc. Likewise check the path and switch (1 or 0) for other libraries. OpenMPI in Ubuntu is for example found in /usr/lib/openmpi and consider experimenting with Atlas later.

Launching the install-script

Once done with the preparation, enter the Install/ folder and type python . Lean back and say a silent prayer. The script should now build and install the required modules into /home/user/code_saturne


   • Consult the readme.txt (Install/) if you encounter problems -
      or visit the forum on - It is the most active Code_saturne
      forum at the moment
   • I had to install extra latex-packages during the install-script
   • After installing is done and you sourced the cs_profile correctly,
      you should be able to create a study with cs_create -study studyname
      - otherwise return to start, rinse and repeat. Most errors are
      (somewhat) easily traced, others require the forum at
      - now go and have fun :)

Claus Andersen 2008