I've noticed that I can't just fix this myself, very bad things
happened to the merged communicator, so this is not a trivial fix I
gather.
Michael
On Apr 30, 2006, at 12:16 PM, Michael Kluskens wrote:
MPI_Intercomm_Merge( intercomm, high, newintracomm, ier )
None of the books I have state the variable type of the second
argument for MPI_Intercomm_Merge for Fortran. Class notes I have
from David Cronk state it is a Logical.
In C it is an "int" with values of true and false.
Looking at OpenMPI 1.1a4r9780 (& openmpi-1.2a1r9780) I find the
following:
./ompi/mpi/f77/intercomm_merge_f.c:
OMPI_GENERATE_F77_BINDINGS (PMPI_INTERCOMM_MERGE,
pmpi_intercomm_merge,
pmpi_intercomm_merge_,
pmpi_intercomm_merge__,
pmpi_intercomm_merge_f,
(MPI_Fint *intercomm, MPI_Flogical *high,
MPI_Fint *newintercomm, MPI_Fint *ierr),
(intercomm, high, newintercomm, ierr) )
and
*ierr = MPI_Intercomm_merge (c_intercomm, OMPI_LOGICAL_2_INT
(*high),
&c_newcomm);
This implies to me that in Fortran the second argument should be a
logical which makes sense.
But in ./ompi/mpi/f90/mpi-f90-interfaces.h, I find:
-------------
interface MPI_Intercomm_merge
subroutine MPI_Intercomm_merge(intercomm, high, newintercomm, ierr)
include 'mpif.h'
integer, intent(in) :: intercomm
integer, intent(in) :: high
integer, intent(out) :: newintercomm
integer, intent(out) :: ierr
end subroutine MPI_Intercomm_merge
end interface MPI_Intercomm_merge
-------------
which I suspect is wrong or at least incomplete with regards to the
second argument.
In which case the problem is also in ./ompi/mpi/f90/scripts/mpi-f90-
interfaces.h.sh
----------
subroutine ${procedure}(intercomm, high, newintercomm, ierr)
include 'mpif.h'
integer, intent(in) :: intercomm
integer, intent(in) :: high
integer, intent(out) :: newintercomm
integer, intent(out) :: ierr
end subroutine ${procedure}
EOF
}
start MPI_Intercomm_merge small
output_162 MPI_Intercomm_merge
end MPI_Intercomm_merge
-------------
Michael
