Hello,
Running Amber in both serial and parallel is quite possible on Mac OS
X, though there are a couple things you'll have to take care of.
First thing I would check is to make sure that the compilers specified
in config.h are mpif77/mpif90 (either can be used if you have both,
though both should point to gfortran, so I'll use mpif77 throughout
the rest of the email) for your fortran compiler and mpicc for your c
compiler. You can check what these compilers are by the commands
"mpif77 -show" and "mpicc -show". The default openMPI included with
xcode does not have fortran support. You can specifically invoke
mpif77 and mpicc from previous installations by using their full path
(and don't forget to set MPI_HOME to the installation you want. It
should be set to the location of mpirun minus the 'bin/mpirun' at the
end. i.e. if your lam mpirun is located in /usr/local/bin/mpirun, then
MPI_HOME is /usr/local). If this doesn't work, see below:
If you are using Snow Leopard, you'll have to get either the gfortran
available on
http://hpc.sourceforge.net/ for snow leopard or use
Macports/Fink to install a 64-bit compatible gcc (4.4 is a stable
release, 4.5 is beta).
Next, Xcode Tools comes with a (broken) version of OpenMPI already.
It is broken in the fact that it was not configured or built with
fortran compiler support. This installation is in /usr/bin, whereas
custom installations usually dump into /usr/local. Therefore, if you
make no changes to your environment, your PATH hierarchy will
typically pull from /usr/ before /usr/local.
The first thing you should do is to uninstall all of your
custom-installed MPIs (since they're obviously not working. I think
it's best to start with a clean system). If you still have the source
code somewhere, just type "make uninstall". You may have to give root
privileges.
Before installing, make sure you have all compilers in your path (this
mainly applies to gfortran, as gcc will already be there if you've
installed xcode). I built mine parallel Amber just fine with both
OpenMPI and MPICH2, though my personal preference goes to MPICH2.
Download the MIPCH2 source code and run configure, pointing to /usr as
the install directory. You can do this with the following command:
./configure --prefix=/usr
Then, type "make" and "sudo make install" afterwards. This should
successfully install MPICH2 in /usr (MPI libraries in /usr/lib, MPI
executables in /usr/bin, and MPI include files in /usr/include).
Check the installation by using mpirun --version, which will give an
error about no mpd process running, and tell you to look at the MPICH2
user guide if you're successful. You can then check to make sure that
both fortran and c compilers have been successfully installed by using
the commands:
mpicc -show
mpif77 -show
You should hopefully get the following output:
gcc -I/usr/include -L/usr/lib -L/usr/lib -lpmpich -lmpich -lopa -lpthread
gfortran -I/usr/include -L/usr/lib -L/usr/lib -lpmpich -lmpich -lopa -lpthread
After that, installation of amber in parallel should be fairly
straight-forward. If your installation still fails, you'd have to
post back here with specific error messages so that we can help. All
I did here was explain in detail the steps I took to install amber in
parallel on my system.
Good luck!
Jason
On Sun, Feb 28, 2010 at 7:15 AM, Christine Baranski
<chbaranski.googlemail.com> wrote:
> Is installation of parallel amber on macbook pro is possible?
>
> I have tried with different combination of mpi and compilers (ifort, g95,
> gfortran) but non of these worked for me. I have also tried installing
> installing lates lam-mpi but still was not able to install a parallel amber.
>
> In addition I have tried mpich, openmpi via fink and openmpi from source.
>
> If someone have installed amber in parallel on macbook pro please guide me
> which combination of mpi and compilers to us?
>
> Regards
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
--
---------------------------------------
Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Graduate Student
352-392-4032
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sun Feb 28 2010 - 09:30:02 PST
