15 |
this section, we provide information on how to customize the code when |
this section, we provide information on how to customize the code when |
16 |
you are ready to try implementing the configuration you have in mind. |
you are ready to try implementing the configuration you have in mind. |
17 |
|
|
|
|
|
18 |
\section{Where to find information} |
\section{Where to find information} |
19 |
\label{sect:whereToFindInfo} |
\label{sect:whereToFindInfo} |
20 |
\begin{rawhtml} |
\begin{rawhtml} |
602 |
where we are re-directing the stream of text output to the file |
where we are re-directing the stream of text output to the file |
603 |
\texttt{output.txt}. |
\texttt{output.txt}. |
604 |
|
|
605 |
|
\subsection{Building/compiling the code elsewhere} |
606 |
|
|
607 |
|
In the example above (section \ref{sect:buildingCode}) we built the |
608 |
|
executable in the {\em input} directory of the experiment for |
609 |
|
convenience. You can also configure and compile the code in other |
610 |
|
locations, for example on a scratch disk with out having to copy the |
611 |
|
entire source tree. The only requirement to do so is you have {\tt |
612 |
|
genmake2} in your path or you know the absolute path to {\tt |
613 |
|
genmake2}. |
614 |
|
|
615 |
|
The following sections outline some possible methods of organizing |
616 |
|
your source and data. |
617 |
|
|
618 |
|
\subsubsection{Building from the {\em ../code directory}} |
619 |
|
|
620 |
|
This is just as simple as building in the {\em input/} directory: |
621 |
|
\begin{verbatim} |
622 |
|
% cd verification/exp2/code |
623 |
|
% ../../../tools/genmake2 |
624 |
|
% make depend |
625 |
|
% make |
626 |
|
\end{verbatim} |
627 |
|
However, to run the model the executable ({\em mitgcmuv}) and input |
628 |
|
files must be in the same place. If you only have one calculation to make: |
629 |
|
\begin{verbatim} |
630 |
|
% cd ../input |
631 |
|
% cp ../code/mitgcmuv ./ |
632 |
|
% ./mitgcmuv > output.txt |
633 |
|
\end{verbatim} |
634 |
|
or if you will be making multiple runs with the same executable: |
635 |
|
\begin{verbatim} |
636 |
|
% cd ../ |
637 |
|
% cp -r input run1 |
638 |
|
% cp code/mitgcmuv run1 |
639 |
|
% cd run1 |
640 |
|
% ./mitgcmuv > output.txt |
641 |
|
\end{verbatim} |
642 |
|
|
643 |
|
\subsubsection{Building from a new directory} |
644 |
|
|
645 |
|
Since the {\em input} directory contains input files it is often more |
646 |
|
useful to keep {\em input} pristine and build in a new directory |
647 |
|
within {\em verification/exp2/}: |
648 |
|
\begin{verbatim} |
649 |
|
% cd verification/exp2 |
650 |
|
% mkdir build |
651 |
|
% cd build |
652 |
|
% ../../../tools/genmake2 -mods=../code |
653 |
|
% make depend |
654 |
|
% make |
655 |
|
\end{verbatim} |
656 |
|
This builds the code exactly as before but this time you need to copy |
657 |
|
either the executable or the input files or both in order to run the |
658 |
|
model. For example, |
659 |
|
\begin{verbatim} |
660 |
|
% cp ../input/* ./ |
661 |
|
% ./mitgcmuv > output.txt |
662 |
|
\end{verbatim} |
663 |
|
or if you tend to make multiple runs with the same executable then |
664 |
|
running in a new directory each time might be more appropriate: |
665 |
|
\begin{verbatim} |
666 |
|
% cd ../ |
667 |
|
% mkdir run1 |
668 |
|
% cp build/mitgcmuv run1/ |
669 |
|
% cp input/* run1/ |
670 |
|
% cd run1 |
671 |
|
% ./mitgcmuv > output.txt |
672 |
|
\end{verbatim} |
673 |
|
|
674 |
|
\subsubsection{Building on a scratch disk} |
675 |
|
|
676 |
|
Model object files and output data can use up large amounts of disk |
677 |
|
space so it is often the case that you will be operating on a large |
678 |
|
scratch disk. Assuming the model source is in {\em ~/MITgcm} then the |
679 |
|
following commands will build the model in {\em /scratch/exp2-run1}: |
680 |
|
\begin{verbatim} |
681 |
|
% cd /scratch/exp2-run1 |
682 |
|
% ~/MITgcm/tools/genmake2 -rootdir=~/MITgcm \ |
683 |
|
-mods=~/MITgcm/verification/exp2/code |
684 |
|
% make depend |
685 |
|
% make |
686 |
|
\end{verbatim} |
687 |
|
To run the model here, you'll need the input files: |
688 |
|
\begin{verbatim} |
689 |
|
% cp ~/MITgcm/verification/exp2/input/* ./ |
690 |
|
% ./mitgcmuv > output.txt |
691 |
|
\end{verbatim} |
692 |
|
|
693 |
|
As before, you could build in one directory and make multiple runs of |
694 |
|
the one experiment: |
695 |
|
\begin{verbatim} |
696 |
|
% cd /scratch/exp2 |
697 |
|
% mkdir build |
698 |
|
% cd build |
699 |
|
% ~/MITgcm/tools/genmake2 -rootdir=~/MITgcm \ |
700 |
|
-mods=~/MITgcm/verification/exp2/code |
701 |
|
% make depend |
702 |
|
% make |
703 |
|
% cd ../ |
704 |
|
% cp -r ~/MITgcm/verification/exp2/input run2 |
705 |
|
% cd run2 |
706 |
|
% ./mitgcmuv > output.txt |
707 |
|
\end{verbatim} |
708 |
|
|
709 |
|
|
710 |
|
\subsection{Using \texttt{genmake2}} |
711 |
|
\label{sect:genmake} |
712 |
|
|
713 |
|
To compile the code, first use the program \texttt{genmake2} (located |
714 |
|
in the \texttt{tools} directory) to generate a Makefile. |
715 |
|
\texttt{genmake2} is a shell script written to work with all |
716 |
|
``sh''--compatible shells including bash v1, bash v2, and Bourne. |
717 |
|
Internally, \texttt{genmake2} determines the locations of needed |
718 |
|
files, the compiler, compiler options, libraries, and Unix tools. It |
719 |
|
relies upon a number of ``optfiles'' located in the |
720 |
|
\texttt{tools/build\_options} directory. |
721 |
|
|
722 |
|
The purpose of the optfiles is to provide all the compilation options |
723 |
|
for particular ``platforms'' (where ``platform'' roughly means the |
724 |
|
combination of the hardware and the compiler) and code configurations. |
725 |
|
Given the combinations of possible compilers and library dependencies |
726 |
|
({\it eg.} MPI and NetCDF) there may be numerous optfiles available |
727 |
|
for a single machine. The naming scheme for the majority of the |
728 |
|
optfiles shipped with the code is |
729 |
|
\begin{center} |
730 |
|
{\bf OS\_HARDWARE\_COMPILER } |
731 |
|
\end{center} |
732 |
|
where |
733 |
|
\begin{description} |
734 |
|
\item[OS] is the name of the operating system (generally the |
735 |
|
lower-case output of the {\tt 'uname'} command) |
736 |
|
\item[HARDWARE] is a string that describes the CPU type and |
737 |
|
corresponds to output from the {\tt 'uname -m'} command: |
738 |
|
\begin{description} |
739 |
|
\item[ia32] is for ``x86'' machines such as i386, i486, i586, i686, |
740 |
|
and athlon |
741 |
|
\item[ia64] is for Intel IA64 systems (eg. Itanium, Itanium2) |
742 |
|
\item[amd64] is AMD x86\_64 systems |
743 |
|
\item[ppc] is for Mac PowerPC systems |
744 |
|
\end{description} |
745 |
|
\item[COMPILER] is the compiler name (generally, the name of the |
746 |
|
FORTRAN executable) |
747 |
|
\end{description} |
748 |
|
|
749 |
|
In many cases, the default optfiles are sufficient and will result in |
750 |
|
usable Makefiles. However, for some machines or code configurations, |
751 |
|
new ``optfiles'' must be written. To create a new optfile, it is |
752 |
|
generally best to start with one of the defaults and modify it to suit |
753 |
|
your needs. Like \texttt{genmake2}, the optfiles are all written |
754 |
|
using a simple ``sh''--compatible syntax. While nearly all variables |
755 |
|
used within \texttt{genmake2} may be specified in the optfiles, the |
756 |
|
critical ones that should be defined are: |
757 |
|
|
758 |
|
\begin{description} |
759 |
|
\item[FC] the FORTRAN compiler (executable) to use |
760 |
|
\item[DEFINES] the command-line DEFINE options passed to the compiler |
761 |
|
\item[CPP] the C pre-processor to use |
762 |
|
\item[NOOPTFLAGS] options flags for special files that should not be |
763 |
|
optimized |
764 |
|
\end{description} |
765 |
|
|
766 |
|
For example, the optfile for a typical Red Hat Linux machine (``ia32'' |
767 |
|
architecture) using the GCC (g77) compiler is |
768 |
|
\begin{verbatim} |
769 |
|
FC=g77 |
770 |
|
DEFINES='-D_BYTESWAPIO -DWORDLENGTH=4' |
771 |
|
CPP='cpp -traditional -P' |
772 |
|
NOOPTFLAGS='-O0' |
773 |
|
# For IEEE, use the "-ffloat-store" option |
774 |
|
if test "x$IEEE" = x ; then |
775 |
|
FFLAGS='-Wimplicit -Wunused -Wuninitialized' |
776 |
|
FOPTIM='-O3 -malign-double -funroll-loops' |
777 |
|
else |
778 |
|
FFLAGS='-Wimplicit -Wunused -ffloat-store' |
779 |
|
FOPTIM='-O0 -malign-double' |
780 |
|
fi |
781 |
|
\end{verbatim} |
782 |
|
|
783 |
|
If you write an optfile for an unrepresented machine or compiler, you |
784 |
|
are strongly encouraged to submit the optfile to the MITgcm project |
785 |
|
for inclusion. Please send the file to the |
786 |
|
\begin{rawhtml} <A href="mail-to:MITgcm-support@mitgcm.org"> \end{rawhtml} |
787 |
|
\begin{center} |
788 |
|
MITgcm-support@mitgcm.org |
789 |
|
\end{center} |
790 |
|
\begin{rawhtml} </A> \end{rawhtml} |
791 |
|
mailing list. |
792 |
|
|
793 |
|
In addition to the optfiles, \texttt{genmake2} supports a number of |
794 |
|
helpful command-line options. A complete list of these options can be |
795 |
|
obtained from: |
796 |
|
\begin{verbatim} |
797 |
|
% genmake2 -h |
798 |
|
\end{verbatim} |
799 |
|
|
800 |
|
The most important command-line options are: |
801 |
|
\begin{description} |
802 |
|
|
803 |
|
\item[\texttt{--optfile=/PATH/FILENAME}] specifies the optfile that |
804 |
|
should be used for a particular build. |
805 |
|
|
806 |
|
If no "optfile" is specified (either through the command line or the |
807 |
|
MITGCM\_OPTFILE environment variable), genmake2 will try to make a |
808 |
|
reasonable guess from the list provided in {\em |
809 |
|
tools/build\_options}. The method used for making this guess is |
810 |
|
to first determine the combination of operating system and hardware |
811 |
|
(eg. "linux\_ia32") and then find a working FORTRAN compiler within |
812 |
|
the user's path. When these three items have been identified, |
813 |
|
genmake2 will try to find an optfile that has a matching name. |
814 |
|
|
815 |
|
\item[\texttt{--pdefault='PKG1 PKG2 PKG3 ...'}] specifies the default |
816 |
|
set of packages to be used. The normal order of precedence for |
817 |
|
packages is as follows: |
818 |
|
\begin{enumerate} |
819 |
|
\item If available, the command line (\texttt{--pdefault}) settings |
820 |
|
over-rule any others. |
821 |
|
|
822 |
|
\item Next, \texttt{genmake2} will look for a file named |
823 |
|
``\texttt{packages.conf}'' in the local directory or in any of the |
824 |
|
directories specified with the \texttt{--mods} option. |
825 |
|
|
826 |
|
\item Finally, if neither of the above are available, |
827 |
|
\texttt{genmake2} will use the \texttt{/pkg/pkg\_default} file. |
828 |
|
\end{enumerate} |
829 |
|
|
830 |
|
\item[\texttt{--pdepend=/PATH/FILENAME}] specifies the dependency file |
831 |
|
used for packages. |
832 |
|
|
833 |
|
If not specified, the default dependency file {\em pkg/pkg\_depend} |
834 |
|
is used. The syntax for this file is parsed on a line-by-line basis |
835 |
|
where each line containes either a comment ("\#") or a simple |
836 |
|
"PKGNAME1 (+|-)PKGNAME2" pairwise rule where the "+" or "-" symbol |
837 |
|
specifies a "must be used with" or a "must not be used with" |
838 |
|
relationship, respectively. If no rule is specified, then it is |
839 |
|
assumed that the two packages are compatible and will function |
840 |
|
either with or without each other. |
841 |
|
|
842 |
|
\item[\texttt{--adof=/path/to/file}] specifies the "adjoint" or |
843 |
|
automatic differentiation options file to be used. The file is |
844 |
|
analogous to the ``optfile'' defined above but it specifies |
845 |
|
information for the AD build process. |
846 |
|
|
847 |
|
The default file is located in {\em |
848 |
|
tools/adjoint\_options/adjoint\_default} and it defines the "TAF" |
849 |
|
and "TAMC" compilers. An alternate version is also available at |
850 |
|
{\em tools/adjoint\_options/adjoint\_staf} that selects the newer |
851 |
|
"STAF" compiler. As with any compilers, it is helpful to have their |
852 |
|
directories listed in your {\tt \$PATH} environment variable. |
853 |
|
|
854 |
|
\item[\texttt{--mods='DIR1 DIR2 DIR3 ...'}] specifies a list of |
855 |
|
directories containing ``modifications''. These directories contain |
856 |
|
files with names that may (or may not) exist in the main MITgcm |
857 |
|
source tree but will be overridden by any identically-named sources |
858 |
|
within the ``MODS'' directories. |
859 |
|
|
860 |
|
The order of precedence for this "name-hiding" is as follows: |
861 |
|
\begin{itemize} |
862 |
|
\item ``MODS'' directories (in the order given) |
863 |
|
\item Packages either explicitly specified or provided by default |
864 |
|
(in the order given) |
865 |
|
\item Packages included due to package dependencies (in the order |
866 |
|
that that package dependencies are parsed) |
867 |
|
\item The "standard dirs" (which may have been specified by the |
868 |
|
``-standarddirs'' option) |
869 |
|
\end{itemize} |
870 |
|
|
871 |
|
\item[\texttt{--mpi}] This option enables certain MPI features (using |
872 |
|
CPP \texttt{\#define}s) within the code and is necessary for MPI |
873 |
|
builds (see Section \ref{sect:mpi-build}). |
874 |
|
|
875 |
|
\item[\texttt{--make=/path/to/gmake}] Due to the poor handling of |
876 |
|
soft-links and other bugs common with the \texttt{make} versions |
877 |
|
provided by commercial Unix vendors, GNU \texttt{make} (sometimes |
878 |
|
called \texttt{gmake}) should be preferred. This option provides a |
879 |
|
means for specifying the make executable to be used. |
880 |
|
|
881 |
|
\item[\texttt{--bash=/path/to/sh}] On some (usually older UNIX) |
882 |
|
machines, the ``bash'' shell is unavailable. To run on these |
883 |
|
systems, \texttt{genmake2} can be invoked using an ``sh'' (that is, |
884 |
|
a Bourne, POSIX, or compatible) shell. The syntax in these |
885 |
|
circumstances is: |
886 |
|
\begin{center} |
887 |
|
\texttt{\% /bin/sh genmake2 -bash=/bin/sh [...options...]} |
888 |
|
\end{center} |
889 |
|
where \texttt{/bin/sh} can be replaced with the full path and name |
890 |
|
of the desired shell. |
891 |
|
|
892 |
|
\end{description} |
893 |
|
|
894 |
|
|
895 |
|
\subsection{Building with MPI} |
896 |
|
\label{sect:mpi-build} |
897 |
|
|
898 |
|
Building MITgcm to use MPI libraries can be complicated due to the |
899 |
|
variety of different MPI implementations available, their dependencies |
900 |
|
or interactions with different compilers, and their often ad-hoc |
901 |
|
locations within file systems. For these reasons, its generally a |
902 |
|
good idea to start by finding and reading the documentation for your |
903 |
|
machine(s) and, if necessary, seeking help from your local systems |
904 |
|
administrator. |
905 |
|
|
906 |
|
The steps for building MITgcm with MPI support are: |
907 |
|
\begin{enumerate} |
908 |
|
|
909 |
|
\item Determine the locations of your MPI-enabled compiler and/or MPI |
910 |
|
libraries and put them into an options file as described in Section |
911 |
|
\ref{sect:genmake}. One can start with one of the examples in: |
912 |
|
\begin{rawhtml} <A |
913 |
|
href="http://mitgcm.org/cgi-bin/viewcvs.cgi/MITgcm/tools/build_options/"> |
914 |
|
\end{rawhtml} |
915 |
|
\begin{center} |
916 |
|
\texttt{MITgcm/tools/build\_options/} |
917 |
|
\end{center} |
918 |
|
\begin{rawhtml} </A> \end{rawhtml} |
919 |
|
such as \texttt{linux\_ia32\_g77+mpi\_cg01} or |
920 |
|
\texttt{linux\_ia64\_efc+mpi} and then edit it to suit the machine at |
921 |
|
hand. You may need help from your user guide or local systems |
922 |
|
administrator to determine the exact location of the MPI libraries. |
923 |
|
If libraries are not installed, MPI implementations and related |
924 |
|
tools are available including: |
925 |
|
\begin{itemize} |
926 |
|
\item \begin{rawhtml} <A |
927 |
|
href="http://www-unix.mcs.anl.gov/mpi/mpich/"> |
928 |
|
\end{rawhtml} |
929 |
|
MPICH |
930 |
|
\begin{rawhtml} </A> \end{rawhtml} |
931 |
|
|
932 |
|
\item \begin{rawhtml} <A |
933 |
|
href="http://www.lam-mpi.org/"> |
934 |
|
\end{rawhtml} |
935 |
|
LAM/MPI |
936 |
|
\begin{rawhtml} </A> \end{rawhtml} |
937 |
|
|
938 |
|
\item \begin{rawhtml} <A |
939 |
|
href="http://www.osc.edu/~pw/mpiexec/"> |
940 |
|
\end{rawhtml} |
941 |
|
MPIexec |
942 |
|
\begin{rawhtml} </A> \end{rawhtml} |
943 |
|
\end{itemize} |
944 |
|
|
945 |
|
\item Build the code with the \texttt{genmake2} \texttt{-mpi} option |
946 |
|
(see Section \ref{sect:genmake}) using commands such as: |
947 |
|
{\footnotesize \begin{verbatim} |
948 |
|
% ../../../tools/genmake2 -mods=../code -mpi -of=YOUR_OPTFILE |
949 |
|
% make depend |
950 |
|
% make |
951 |
|
\end{verbatim} } |
952 |
|
|
953 |
|
\item Run the code with the appropriate MPI ``run'' or ``exec'' |
954 |
|
program provided with your particular implementation of MPI. |
955 |
|
Typical MPI packages such as MPICH will use something like: |
956 |
|
\begin{verbatim} |
957 |
|
% mpirun -np 4 -machinefile mf ./mitgcmuv |
958 |
|
\end{verbatim} |
959 |
|
Sightly more complicated scripts may be needed for many machines |
960 |
|
since execution of the code may be controlled by both the MPI |
961 |
|
library and a job scheduling and queueing system such as PBS, |
962 |
|
LoadLeveller, Condor, or any of a number of similar tools. A few |
963 |
|
example scripts (those used for our \begin{rawhtml} <A |
964 |
|
href="http://mitgcm.org/testing.html"> \end{rawhtml}regular |
965 |
|
verification runs\begin{rawhtml} </A> \end{rawhtml}) are available |
966 |
|
at: |
967 |
|
\begin{rawhtml} <A |
968 |
|
href="http://mitgcm.org/cgi-bin/viewcvs.cgi/MITgcm_contrib/test_scripts/"> |
969 |
|
\end{rawhtml} |
970 |
|
{\footnotesize \tt |
971 |
|
http://mitgcm.org/cgi-bin/viewcvs.cgi/MITgcm\_contrib/test\_scripts/ } |
972 |
|
\begin{rawhtml} </A> \end{rawhtml} |
973 |
|
|
974 |
|
\end{enumerate} |
975 |
|
|
976 |
|
An example of the above process on the MITgcm cluster (``cg01'') using |
977 |
|
the GNU g77 compiler and the mpich MPI library is: |
978 |
|
|
979 |
|
{\footnotesize \begin{verbatim} |
980 |
|
% cd MITgcm/verification/exp5 |
981 |
|
% mkdir build |
982 |
|
% cd build |
983 |
|
% ../../../tools/genmake2 -mpi -mods=../code \ |
984 |
|
-of=../../../tools/build_options/linux_ia32_g77+mpi_cg01 |
985 |
|
% make depend |
986 |
|
% make |
987 |
|
% cd ../input |
988 |
|
% /usr/local/pkg/mpi/mpi-1.2.4..8a-gm-1.5/g77/bin/mpirun.ch_gm \ |
989 |
|
-machinefile mf --gm-kill 5 -v -np 2 ../build/mitgcmuv |
990 |
|
\end{verbatim} } |
991 |
|
|
992 |
\section[Running MITgcm]{Running the model in prognostic mode} |
\section[Running MITgcm]{Running the model in prognostic mode} |
993 |
\label{sect:runModel} |
\label{sect:runModel} |