mirror of
https://github.com/nwchemgit/nwchem.git
synced 2026-07-21 06:25:21 -04:00
36c95e79c8
DEC OS/F is an obsolete Unix operating system (https://en.wikipedia.org/wiki/Tru64_UNIX).
200 lines
8.6 KiB
TeX
200 lines
8.6 KiB
TeX
\label{sec:install}
|
|
|
|
This chapter contains guidance on how to obtain a copy of NWChem and
|
|
install it on your system. The best source for installation instructions
|
|
is the INSTALL file in the NWChem source distribution, so those
|
|
instructions will not be repeated here. If you have problems with the
|
|
installation, you can request help from NWChem
|
|
support via e-mail at {\tt nwchem-support@emsl.pnl.gov}.
|
|
|
|
The following subsections discuss some of the important considerations
|
|
when installing NWChem, and provide information on environmental
|
|
variables, libraries, and makefiles needed to run the code.
|
|
|
|
\section{How to Obtain NWChem}
|
|
|
|
The NWChem source code tree current release is version 5.1.1. To obtain NWChem
|
|
a User's Agreement must be properly filled out and sent to us. The User's
|
|
Agreement may be found on the NWChem webpages at
|
|
|
|
\begin{verbatim}
|
|
http://www.emsl.pnl.gov:2080/docs/nwchem
|
|
\end{verbatim}
|
|
|
|
by clicking on the link "Download" and following the
|
|
instructions as they appear. If you already have an older version of NWChem,
|
|
new download informaiton may be obtained at the location on the web.
|
|
If you have any problems
|
|
using the WWW pages or forms, or getting access to the code, send e-mail to
|
|
{\tt nwchem-support@emsl.pnl.gov}.
|
|
|
|
\section{Supported Platforms}
|
|
\label{sec:platforms}
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
% NOTE: this section is adapted from the current (as of 3/28/01) version of
|
|
% the script INSTALL for NWChem, in the CVS repository. If INSTALL
|
|
% has been updated since, this section should be updated, too.
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
NWChem is readily portable to essentially any sequential or parallel computer.
|
|
The source code currently contains options for versions that will run
|
|
on the following platforms.
|
|
|
|
\begin{verbatim}
|
|
|
|
NWCHEM_TARGET Platform Checked OS/Version Precision
|
|
-----------------------------------------------------------------
|
|
SOLARIS Sun *** Solaris 2.X double
|
|
IBM IBM RS/6000 *** AIX 4.X double
|
|
SGI_N32 SGI 64 bit os *** IRIX 6.5 double
|
|
using 32 ints
|
|
SGITFP SGI 64 bit os *** IRIX 6.5 double
|
|
CRAY-T3D Cray T3D UNICOS single
|
|
CRAY-T3E Cray T3E *** UNICOS single
|
|
LAPI IBM SP *** AIX/LAPI double
|
|
LINUX Intel x86 *** RedHat 5.2-6.2 double
|
|
PowerPC ** RedHat 6.0 double
|
|
LINUX64 Alpha ** RedHat 6.2 double
|
|
HPUX HP ** HPUX 11.0 double
|
|
WIN32 Intel x86 * Windows98/NT double
|
|
-----------------------------------------------------------------
|
|
*Note: LAPI is now the primary way to use NWChem on an IBM SP system.
|
|
If you don't have it get it from IBM.
|
|
|
|
\end{verbatim}
|
|
The environment variable {\tt NWCHEM\_TARGET} must
|
|
be set to the symbolic name
|
|
that matches your target platform. For example, if you are installing
|
|
the code on an IBM SP, the command is
|
|
|
|
\begin{verbatim}
|
|
% setenv NWCHEM_TARGET LAPI
|
|
\end{verbatim}
|
|
|
|
Refer to Section \ref{sec:envar} for additional discussion of environmental variables
|
|
required by NWChem.
|
|
|
|
\subsection{Porting Notes}
|
|
\label{sec:PortingNotes}
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
% NOTE: this section is adapted from the current (as of 9/29/98) version of
|
|
% the file Porting.notes, from ~/doc/ in the CVS repository. If Porting.notes
|
|
% has been updated since, this section should be updated, too.
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
|
|
While it is true that NWChem will run on {\em almost} any computer, there are always
|
|
a few jokers in the deck. Here are some that have been found, and were
|
|
considered sufficiently amusing to be documented.
|
|
|
|
\begin{itemize}
|
|
\item from the Intel Paragon OSF/1 R1.2.1 (discovered 16 July 1994 by DE Bernholdt);
|
|
PGI's compilation system is braindamaged in some fascinating ways:
|
|
\begin{enumerate}
|
|
\item cpp860 by default defines {\tt \_\_PARAGON\_\_} and other things, as stated in
|
|
the {\tt man} page, but when invoked by {\tt if77}, these things are {\em not} defined.
|
|
\item ld's -L prepends directories to the search path instead of
|
|
appending, as is done in almost every other unix compiler package
|
|
\end{enumerate}
|
|
|
|
\item from the HP-UX 9000/735, also some others (reported 08 Feb 1996 by Jarek Nieplocha):
|
|
|
|
\begin{enumerate}
|
|
\item Avoid the "free" HP C compiler - use gcc instead:
|
|
HP cc does not generate any symbols or code for several routines in one of
|
|
the GA files. To make the user's life more entertaining, there are no
|
|
warning or error messages either -- compiler creates a junk object file
|
|
quietly and pretends that everything went well.
|
|
(Karl Anderson says: "(HP) cc is worth every penny you paid for it.")
|
|
|
|
\item {\tt fort77} instead of {\tt f77} should be used to link fortran programs, since
|
|
{\tt f77} doesn't support the {\tt -L} flag. Fortran code should be compiled with the
|
|
{\tt +ppu} flag that adds underscores to the subroutine names.
|
|
\end{enumerate}
|
|
|
|
\end{itemize}
|
|
|
|
|
|
\section{Environmental Variables}
|
|
\label{sec:envar}
|
|
|
|
There are mandatory environmental variables, as well as optional ones,
|
|
that need to be set for the compilation of NWChem to work correctly.
|
|
The mandatory one are listed first:
|
|
|
|
\begin{table}[htbp]
|
|
\begin{center}
|
|
\begin{tabular}{lcc}
|
|
\verb+NWCHEM_TOP+ & the top directory of the NWChem tree, e.g.\\
|
|
\verb+ setenv NWCHEM_TOP /u/adrian/nwchem+\\
|
|
\\
|
|
\verb+NWCHEM_TARGET+ & the symbolic name that matches your target
|
|
\\
|
|
& platform, e.g.\\
|
|
\verb+ setenv NWCHEM_TARGET LAPI+\\
|
|
\\
|
|
\verb+NWCHEM_MODUELS+ & the modules you want included in the binary
|
|
\\
|
|
& that you build, e.g.\\
|
|
\verb+ setenv NWCHEM_MODULES "all gapss"+
|
|
\end{tabular}
|
|
\end{center}
|
|
\end{table}
|
|
|
|
The following environment variables which tell NWChem more about your
|
|
system are optional. If they are not set, NWChem will try to pick
|
|
reasonable defaults:
|
|
|
|
\begin{table}[htbp]
|
|
\begin{center}
|
|
\begin{tabular}{lcc}
|
|
\verb+NWCHEM_TARGET_CPU+ & more information about a particular\\
|
|
& architechture\\
|
|
\verb+ setenv NWCHEM_TARGET_CPU P2SC+\\
|
|
\\
|
|
\verb+SCRATCH_DEF_DIR+ & default scratch directory for\\
|
|
& temporary files, e.g.\\
|
|
\verb+ setenv SCRATCH_DEF_DIR "\'/scratch\'"+\\
|
|
\\
|
|
\verb+PERMANENT_DEF_DIR+ & default permanent directory for\\
|
|
& files to keep, e.g.\\
|
|
\verb+ setenv PERMANENT_DEF_DIR "\'/home/user\'"+\\
|
|
\\
|
|
\verb+NWCHEM_BASIS_LIBRARY_PATH+ & location of the basis set libraries \\
|
|
& (the builder is responsible to make \\
|
|
& sure that the library gets to the \\
|
|
& place), e.g.\\
|
|
\verb+ setenv NWCHEM_BASIS_LIBRARY_PATH "/bin/libraries/"+\\
|
|
\\
|
|
\verb+LARGE_FILES+ & needed to circumvent the 2 GB limit\\
|
|
& on IBM (note that your system \\
|
|
& administrator must also enable \\
|
|
& large files in the file system), e.g.\\
|
|
\verb+ setenv LARGE_FILES TRUE+\\
|
|
\\
|
|
\verb+JOBTIME_PATH+ & directory where jobtime and jobtime.pl\\
|
|
& will be placed by the builder on \\
|
|
& IBM SP, e.g.\\
|
|
\verb+ setenv JOBTIME_PATH /u/nwchem/bin+\\
|
|
\\
|
|
\verb+LIB_DEFINES+ & additional defines for the C \\
|
|
& preprocessor (for both Fortran \\
|
|
& and C), e.g.\\
|
|
\verb+ setenv LIB_DEFINES -DDFLT_TOT_MEM=16777216+\\
|
|
This sets the dynamic memory available for \\
|
|
NWChem to run, where the units are in doubles.\\
|
|
Check out the Section for MEMORY SCRIPT below.\\
|
|
\\
|
|
\verb+TCGRSH+ & alternate path for rsh, it is intended \\
|
|
& to allow usage of ssh in TCGMSG \\
|
|
& (default communication protocol \\
|
|
& for workstation builds).\\
|
|
\verb+ setenv TCGRSH /usr/local/bin/ssh+\\
|
|
\\
|
|
IMPORTANT: ssh should not ask for a password. \\
|
|
In order to do that:\\
|
|
1) On the master node, run "ssh-keygen"\\
|
|
2) For each slave node, \verb+slave_node+,\\
|
|
\verb+ scp ~/.ssh/identity.pub \+ \\
|
|
\verb+ username@slave_node:.ssh/authorized_keys+
|
|
\end{tabular}
|
|
\end{center}
|
|
\end{table}
|