Hi Jason/Dave,
Thank you for your helps. I tried different ways, finally make it works.
Here is the summary for what I did and the fixes:
First my system had openmpi installed when the system setup. When I compile
parallel version, it did not have mpif90. So I downloaded and installed
openmpi to /usr/bin. Then I have all the c c++ and f compiler. But when I
configure parallel amber, I was told that library was not found. I
consulted Red Hat support. I was given an solution: remove what I had
installed. Reinstall openmpi devel version using yum install: yum install
openmpi-devel. Then the configure was successful. Finally I add
*LD_LIBRARY_PATH
to the openmpi library. Now it all works.
Thanks again,
William
*
On Mon, Sep 12, 2011 at 4:37 PM, Jason Swails <jason.swails.gmail.com>wrote:
> Mostly unrelated to the current thread, but possibly useful information?
>
> On Mon, Sep 12, 2011 at 4:34 PM, David A Case <case.biomaps.rutgers.edu
> >wrote:
>
> > You then should
> > set your MPI_HOME environment variable to $AMBERHOME. (Not sure if the
> > last
> > is required, but I think it might be.) Also set your PATH variable to
> > include
> > $AMBERHOME/bin.
> >
>
> MPI_HOME is only necessary for pmemd.cuda.MPI (since the CUDA compiler
> needs
> to know where the MPI headers are).
>
> All the best,
> Jason
>
> --
> Jason M. Swails
> Quantum Theory Project,
> University of Florida
> Ph.D. Candidate
> 352-392-4032
> _______________________________________________
> 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 Tue Sep 13 2011 - 08:00:05 PDT
