Index: /trunk/LMDZ.MARS/doc/run.tex =================================================================== --- /trunk/LMDZ.MARS/doc/run.tex (revision 2205) +++ /trunk/LMDZ.MARS/doc/run.tex (revision 2206) @@ -94,92 +94,6 @@ \section{Installing the model} -There are two distinct ways of installing/compiling the model: -One (nowdays depreciated) using scripts in the {\tt LMDZ.MARS} -directory, and the other (encouraged because one can then also -run the model on parallel computers) in the {\tt LMDZ.COMMON} directory - -\subsection{Settings in LMDZ.MARS (depreciated)} -\begin{description} -%\item {\bf -} Copy the basic model directory LMDZ.MARS to your account -%(the contents of this directory are described in chapter \ref{loc:contenu}). - -\item {\bf -} Set some environment variables needed for the compilation - of the model (it is also possible to set the environment - variables in the {\tt makegcm} script, as explained below): - \begin{description} - \item {\bf LMDGCM} : Path to the directory where you have put the model - (full path).\\ - If using Csh: - \begin{verbatim} - setenv LMDGCM /where/you/put/the/model/LMDZ.MARS - \end{verbatim} - If using Bash: - \begin{verbatim} - export LMDGCM=/where/you/put/the/model/LMDZ.MARS - \end{verbatim} - \item {\bf LIBOGCM} : Path to the directory - ({\tt libo} for example) where intermediate objects will be stored - during the compilation of the model with the {\tt makegcm} script - (if that directory does not - exist then {\tt makegcm} will create it).\\ - If using Csh: - \begin{verbatim} - setenv LIBOGCM /where/you/want/objects/to/go/libo - \end{verbatim} - If using Bash: - \begin{verbatim} - export LIBOGCM=/where/you/want/objects/to/go/libo - \end{verbatim} - \end{description} - -\item {\bf -} Knowing where your NetCDF library is installed, - set environment variables {\bf NCDFINC} and {\bf NCDFLIB}: - - \begin{description} - \item Once the NetCDF library has been compiled (or downloaded), - you should have access to the library {\tt libnetcdf.a} itself, - the various files ({\tt netcdf.inc}, {\tt netcdf.mod}, ...) - to include in programs, and basic NetCDF software ({\it ncdump} - and {\it ncgen}). - - \item To ensure that during compilation, the model can find the - NetCDF library and include files, - you must declare environment variables {\bf NCDFLIB} and {\bf NCDFINC} - (again, it is also possible to set these environment - variables in the {\tt makegcm} script, as explained below). - - \item {\bf NCDFLIB} must contain the path to the directory containing - the object library {\tt libnetcdf.a} - and {\bf NCDFINC} must contain the path to the directory containing - the include files ({\tt netcdf.inc},...)\\ - If using Csh: - \begin{verbatim} - setenv NCDFINC /wherever/is/netcdf/include - setenv NCDFLIB /wherever/is/netcdf/lib - \end{verbatim} - If using Bash: - \begin{verbatim} - export NCDFINC=/wherever/is/netcdf/include - export NCDFLIB=/wherever/is/netcdf/lib - \end{verbatim} - \end{description} - -\item {\bf -} Go to your {\tt LMDZ.MARS} and adapt the {\tt makegcm} - script to fit your needs: - \begin{itemize} - \item Examples of {\tt makegcm} scripts, adapted for different compilers - (pgf90, g95, gfortran and ifort) are provided (files {\tt makegcm}, - {\tt makegcm\_g95}, {\tt makegcm\_gfortran}, {\tt makegcm\_ifort}) copy or - rename the relevant one as {\tt makegcm} in the same directory. - \item As mentionned above, you may edit the script to hard code - values of {\tt LMDGCM}, {\tt LIBOGCM}, {\tt NCDFINC} and {\tt NCDFLIB} - instead of relying on the use of environment variables (see the commented out - examples in the scripts at lignes 20-30). Note that since the {\tt - makegcm} is a Csh script, Csh syntax must be used there. - \end{itemize} - -\end{description} - -\subsection{Settings in LMDZ.COMMON (advised)} +Scripts for installation/compilation for the model are in the {\tt LMDZ.COMMON} directory +These scripts can also run the model on parallel computers. You should first compile the IOIPSL library which is used\footnote{It is in fact for now possible to run the GCM without the IOIPSL library but this requires adding the {\tt -io noioipsl} to the {\tt makelmdz\_fcm} command line, and might no longer be possible in the future.} by the GCM. To do this go to the {\tt LMDZ.COMMON/ioipsl} directory. There are a number of example scripts (depending on machines and compiler suites to use) to run to download and install the ioipsl library. As an illustrative example we detail here using the {\tt install\_ioipsl\_gfortran.bash} script: @@ -192,43 +106,4 @@ \section{Compiling the model} \label{sc:compile} -\subsection{Compiling in LMDZ.MARS (depreciated)} -\noindent As explained above, compiling the model is done -using the {\tt makegcm} script and providing it with the appropriate -options: -\begin{itemize} -\item Example 1: Compiling the Martian model at grid resolution 64x48x25 -for example, type (in compliance with the manual for the makegcm function -given in section~\ref{sc:compil1}) - -\begin{verbatim} -makegcm -d 64x48x25 -p mars gcm -\end{verbatim} - -\noindent -You will find executable {\bf gcm.e} (the compiled model) in the directory -where you ran the makegcm command. - -%{\bf -} Example 2: Compiling the Martian model with 3 tracers -%(e.g. CO2, water vapour and ice to simulate the water cycle): -%\begin{verbatim} -%makegcm -d 64x48x25 -t 2 -p mars gcm -%\end{verbatim} - -\item Example 2: -Compiling the the Martian model with your choice of -compiler options, e.g. to check for array overflow -(useful for debugging: warning, the model is then much slower!): -\begin{verbatim} -makegcm -d 64x48x25 -p mars -O "-C" gcm -\end{verbatim} -Note that the {\tt makegcm} script also has a "debug" option which -includes a collection of adequate debugging options. To use it, -simply add the {\tt -debug} option: -\begin{verbatim} -makegcm -d 64x48x25 -p mars -debug gcm -\end{verbatim} -\end{itemize} - -\subsection{Compiling in LMDZ.COMMON (advised)} The Bash script {\tt makelmdz\_fcm} is used to compile the model. It needs not be modified or adapted to your settings, as all @@ -243,18 +118,17 @@ with the name used for your own arch files), and multiple options: \begin{itemize} -\item Example 1: Compiling the Martian model at grid resolution 64x48x32 -\begin{verbatim} -makelmdz_fcm -arch linux-ifort -d 64x48x29 -p mars gcm -\end{verbatim} -The executable, {\tt gcm\_64x48x29\_phymars\_seq.e} in the present case, -will be generated in the {\tt bin} subdirectory. +\item Example 1: Compiling the Martian model at grid resolution 64x48x49 +\begin{verbatim} +makelmdz_fcm -arch linux-ifort -d 64x48x49 -p mars gcm +\end{verbatim} \item Example 2: Compiling as above but in "debug" mode \begin{verbatim} -makelmdz_fcm -arch linux-ifort -d 64x48x29 -p mars -debug gcm +makelmdz_fcm -arch linux-ifort -d 64x48x49 -p mars -debug gcm \end{verbatim} \item Example 3: Compiling the model to run in parallel (MPI) mode: \begin{verbatim} -makelmdz_fcm -arch linux-ifort -parallel mpi -d 64x48x29 -p mars gcm -\end{verbatim} +makelmdz_fcm -arch linux-ifort -parallel mpi -d 64x48x49 -p mars gcm +\end{verbatim} +This option is different from the -j option that determines the number of cores when compilation is run in parallel mode. \item For an overview of all available options: \begin{verbatim} @@ -323,5 +197,4 @@ \end{verbatim} - You might also want to keep all messages and diagnotics written to standard output (i.e. the screen). You should then redirect the standard output @@ -330,4 +203,11 @@ gcm.e > gcm.out 2>&1 \end{verbatim} + +If you want to use parallel mode (as MPI for instance), you should specify it as follow: +\begin{verbatim} +mpirun gcm.e > gcm.out 2>&1 +\end{verbatim} +NB: The definition of parallel parameters, such as number of cores, is dependant on the machine used. +You should find examples of scripts within other simulations or machine user guides.