Thanks for the reply!
Turns out that what went 'horribly wrong' must have been my
copying-and-pasting...
What I listed as the last lines of output from "./configure_openmpi gnu"
wasn't. (Teaches me to do several things at a rush at the same time and not
checking before pressing 'send')
Sorry.
I ran configure_openmpi again, just in case, and all went fine ("openmpi
configure succeeded."), as did compilation of the .MPI executables
afterwards.
Now I just need to iron out some issues with testing, probably due to the
setup on my local cluster - I don't think I can run in parallel on the
login-node and also have problems when using an interactive job (PBS
command: qsub -I nodes=1:ppn=2).
Thanks,
Marc
On 14 April 2012 04:01, case <case.biomaps.rutgers.edu> wrote:
> On Fri, Apr 13, 2012, Marc van der Kamp wrote:
> >
> > I have downloaded and installed Amber12. I would like to install parallel
> > versions of the executables now (sander.MPI, pmemd.MPI).
> > I was hoping to use the easy script
> > provided: $AMBERHOME/AmberTools/src/configure_openmpi
> > This is what I did:
> >
> > cd $AMBERHOME/AmberTools/src
> > wget
> >
> http://www.open-mpi.org/software/ompi/v1.5/downloads/openmpi-1.5.4.tar.bz2
> > tar xvfj openmpi-1.5.4.tar.bz2
> > ./configure_openmpi gnu
> >
> > I seem to get to the 'make install' stage (final lines of the
> > configure_openmpi script), as these are the final lines from running the
> > above command:
> >
> > rm -f *.o *.obj *.mod pmemd.MPI pmemd pmemd.cuda pmemd.cuda.MPI *.d
> work.pc*
> > `/export/users/chmwvdk/amber12/src/pmemd.amoeba/src'
> > rm -f *.o *.obj *.mod pmemd.amoeba.MPI pmemd.amoeba *.d work.pc*
> > make[2]: Leaving directory
> > `/export/users/chmwvdk/amber12/src/pmemd.amoeba/src'
> > make[1]: Leaving directory
> `/export/users/chmwvdk/amber12/src/pmemd.amoeba'
>
> Looks like something went dreadfully wrong during the configure_openmpi
> script. You need to examine everything that happened there: buried in the
> voluminous output should be a line "openmpi configure succeeded", and a
> line
> about LD_LIBRARY_PATH. Immediately following this should be a line like
> "Making clean in test". Later on you will see hundreds of lines with "CC"
> followed by file names. (The "make install" command in the script should
> be runnning inside amber12/AmberTools/src/openmpi-1.5.4, and wouldn't be
> doing
> anything with any amber files like pmemd.)
>
> Alternatively, there may be a line saying "openmpi configure failed"; at
> this
> point the script should exit, giving you back a command prompt.
>
> Somehow, the script seems to have been running the "make clean" command
> from the amber12 directory. I don't see how this could have happened
> based on what you said you did. It looks like you need to re-run the
> configure_openmpi script, and follow carefully what happens.
>
> >
> > However, I don't have any .MPI executables in $AMBERHOME/bin/ afterwards.
> > AmberTools12.pdf section 1.2 point 7 says: "Follow the instructions of
> > these scripts, then return to step 6."
> > (But step 6 is "make test")
>
> Should say return to step 7. You have to cd to $AMBERHOME and run
> configure
> from there, followed by "make install".
>
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Apr 16 2012 - 02:00:03 PDT
