On Fri, Mar 26, 2010, Jason Swails wrote:
> On Fri, Mar 26, 2010 at 10:40 AM, InSuk Joung <i.joung.gmail.com> wrote:
> > make[2]: Entering directory `/home/isjoung/AMBER/amber11/src/semantics'
> > ./parse *.rules
> > ./parse: error while loading shared libraries: libmpi.so.0: cannot open
> > shared object file: No such file or directory
> > make[2]: *** [checkexpr.c] error 127
> > make[2]: Leaving directory `/home/isjoung/AMBER/amber11/src/semantics'
> > make[1]: *** [checkexpr.c] error 2
> > make[1]: Leaving directory `/home/isjoung/AMBER/amber11/src/nab'
> > make: *** [parallel] error 2
> >
> > There is 'libmpi.so.0' in $AMBERHOME/lib. But 'parse' doesn't seem to
> > search through the directory.
> > After I put it in $LD_LIBRARY_PATH, 'parse' stopped complaining about this.
>
> In my experience, $MPI_HOME/lib has always had to be in my
> LD_LIBRARY_PATH for MPI executables to run successfully. Perhaps we
> should add those directions to the end of the openmpi/lam install
> included with Amber (since it compiles the MPI to $AMBERHOME)? Am I
> mistaken about this?
Good point. I always use -static, so I don't run across those problems.
It would be good to update the docs.
There are still things to learn about the mpi version of NAB, which I've never
used very much. I'm thinking that "make parallel" in the AmberTools side
should just compile ptraj.MPI, and that we tell people to type "make nabonly"
(or some new target) in addition if they want the mpi NAB.
...regards...dac
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Fri Mar 26 2010 - 12:00:02 PDT
