[mpich-discuss] ifort/mpich installation Mac 10.8.5

Thakur, Rajeev thakur at mcs.anl.gov
Wed Jul 16 07:17:15 CDT 2014


Make sure you run the mpirun script also from the MPICH directory. Use the script called mpiexec from there.

Rajeev

On Jul 16, 2014, at 5:43 AM, maurice van putten <mvputten2002 at hotmail.com> wrote:

> Dear Rajeev,
> 
>       Thanks for your reply. I now realize mpich is under /opt/local/include/mpich-mp.
> There is another under /usr/local/include that has an mpif.h with include "mpif-common.h".
> These are by Indiana University/Stuttgart/UC/Cisco.
> 
>  mpif90 -I/opt/local/include/mpich-mp -o tst tst.f  (containing include "mpif.h" in tst.f)
>  mpirun -np 2 ./tst
> 
>  gives a runtime error 
> 
>   1275069467  1275069468  1275070495  (=MPI_INTEGER,MPI_REAL,MPI_DOUBLE_PRECISION)
>   1275069467  1275069468  1275070495
>  *** An error occurred in MPI_Comm_rank
>  *** reported by process [140735101665281,140733193388032]
>  *** on communicator MPI_COMM_WORLD
>  *** MPI_ERR_COMM: invalid communicator
>  *** MPI_ERRORS_ARE_FATAL (processes in this communicator will now abort,
>  ***    and potentially your MPI job)
> (same for 2nd process)
> 
>   Clearly, there I have an initialization error on the basic mpi variables. Is there a possible mix-up 
> between these two mpi packages?
> 
> Best,
> 
> Henry 
> 
> > From: thakur at mcs.anl.gov
> > To: discuss at mpich.org
> > Date: Wed, 16 Jul 2014 09:11:12 +0000
> > Subject: Re: [mpich-discuss] ifort/mpich installation Mac 10.8.5
> > 
> > What is mpif-common.h? You need to include mpif.h and provide the path to the mpich include directory. The problem you are seeing is caused because the wrong mpif.h is being picked up.
> > 
> > Rajeev
> > 
> > 
> > On Jul 16, 2014, at 2:45 AM, maurice van putten <mvputten2002 at hotmail.com> wrote:
> > 
> > > Hi,
> > > 
> > > Perhaps someone can provide a solution to what seems to be
> > > a problem with installation of mpich.
> > > 
> > > I installed ifort (intel) and mpich (via port install), compile
> > > a basic test program mpif90 -I/usr/local/include -o tst tst.f 
> > > (here "I/usr/local/include" is included to find mpif-common.h), 
> > > 
> > > PROGRAM mpitest
> > > integer ierr,my_rank,nm_prcs
> > > include "mpif-common.h"
> > > call MPI_INIT(ierr) 
> > > write(*,*)MPI_INTEGER,MPI_REAL,MPI_DOUBLE_PRECISION
> > > call MPI_COMM_RANK(MPI_COMM_WORLD,my_rank,ierr)
> > > call MPI_COMM_SIZE(MPI_COMM_WORLD,nm_prcs,ierr)
> > > write(*,*)'MPI:',my_rank,nm_prcs,ierr
> > > call MPI_FINALIZE(ierr)
> > > end
> > > 
> > > and do mpirun -np 2 ./tst. The result is
> > > 
> > > 7 13 17
> > > MPI: 0 1 0
> > > 7 13 17
> > > MPI: 0 1 0
> > > 
> > > It shows that nm_prcs = 1 and both processes show the same my_rank = 0,
> > > even though ierr =0. 
> > > 
> > > The initialization in mpirun seem to fail, to assign a proper process count 2.
> > > 
> > > Any ideas on what's amiss?
> > > 
> > > Many thanks,
> > > 
> > > Henry.
> > > _______________________________________________
> > > discuss mailing list discuss at mpich.org
> > > To manage subscription options or unsubscribe:
> > > https://lists.mpich.org/mailman/listinfo/discuss
> > 
> > _______________________________________________
> > discuss mailing list discuss at mpich.org
> > To manage subscription options or unsubscribe:
> > https://lists.mpich.org/mailman/listinfo/discuss
> _______________________________________________
> discuss mailing list     discuss at mpich.org
> To manage subscription options or unsubscribe:
> https://lists.mpich.org/mailman/listinfo/discuss




More information about the discuss mailing list