[mpich-discuss] RE: Buggy code - return status is an Integer array for fort pgms

Jayesh Krishna jayesh at mcs.anl.gov
Tue May 6 10:55:09 CDT 2008


 Hi,
  The return status in MPI calls is an Integer array for fortran programs.
After the change your code works for me.

######### orig code #########

...
integer :: x, y, z, count, k, status, ierr
...
CALL MPI_RECV(count, 1, MPI_INTEGER, source, OCNT_TAG, MPI_COMM_WORLD,
status, ierr)

######### modify to ##########

...
integer :: x, y, z, count, k, status(MPI_STATUS_SIZE), ierr
...
CALL MPI_RECV(count, 1, MPI_INTEGER, source, OCNT_TAG, MPI_COMM_WORLD,
status, ierr)

##############################
 
 Let us know if you have any other problems.

Regards,
Jayesh
-----Original Message-----
From: Gib Bogle [mailto:g.bogle at auckland.ac.nz] 
Sent: Monday, May 05, 2008 2:16 AM
To: Jayesh Krishna
Subject: Buggy code

Hi Jayesh,

After a lot of trial and error, and some luck, I've managed to create a
small program that shows the problem I referred to with
MPICH2/Fortran/Windows.

I've put the source code and the associated Intel project files into a zip
file (attached).

The failure mode is different with 'No optimization' and 'Optimize for
maximum speed'.

This test program runs if I turn off runtime error checking (All), but
that isn't true for the real program.  In general changing the compile
and/or run flags changes the behaviour.

By the way, if the write statement at line 29 in etransfer.f90 is
commented out, the program runs.

My guess is that there is a problem with the stack.

I hope this code is useful to you, and wish you luck.  I'm exhausted :-(

Cheers
Gib

--
Dr. Gib Bogle
Research Scientist
Auckland Bioengineering Institute
University of Auckland
New Zealand

http://www.bioeng.auckland.ac.nz

g.bogle at auckland.ac.nz
(64-9) 373-7599 Ext. 87030
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/mpich-discuss/attachments/20080506/c03113ca/attachment.htm>


More information about the mpich-discuss mailing list