your code is actually not correct. If you look at the MPI specification
you will see that s should also be an array of length nProcs (in your
test), since you send different elements to each process. If you want to
send the same s to each process, you have to use MPI_Bcast.
Thanks
Edgar
Chandan Basu wrote:
I am trying to use MPI_Alltoall in the following program. After
execution all the nodes should show same value for the array su.
However only the root node is showing correct value. other nodes are
giving garbage value. Please help.
I have used openmpi version "1.1.4". The mpif90 uses intel fortran
cbasu
------------------------------------------------
program main
implicit none
include 'mpif.h'
integer :: status(MPI_Status_size)
integer :: ierr, rank, nProcs
double precision :: s
double precision, pointer :: su(:)
call MPI_Init (ierr)
call MPI_Comm_rank(MPI_COMM_WORLD, rank, ierr)
call MPI_Comm_size(MPI_COMM_WORLD, nProcs, ierr)
allocate(su(nProcs))
su = 0.0d0
s = 1.0d0
call MPI_Alltoall(s, 1, MPI_DOUBLE_PRECISION, su, 1, &
& MPI_DOUBLE_PRECISION, MPI_COMM_WORLD, ierr);
! all nodes should have su(1:nProcs) = 1.0 at this pont
print *, rank, su
deallocate(su)
call MPI_Finalize(ierr)
end program main
----------------------------------------------
<http://adworks.rediff.com/cgi-bin/AdWorks/sigclick.cgi/www.rediff.com/signature-home.htm/1507191490@Middle5?PARTNER=3>
------------------------------------------------------------------------
_______________________________________________
users mailing list
us...@open-mpi.org
http://www.open-mpi.org/mailman/listinfo.cgi/users
