[Mpi-forum] An error in mpi program

[wg zhang]

Hi,guys,
I am not sure am I supposed to ask for help here or not. But I do need your
guys' help to give me some hint solving this problem.
I have a parallel fortran code which is written by us. It seems like working
well but actually the output from 1 node to 5 nodes are the same(the results
I think are correct).When I am using more than 6 nodes, the output is
different,which I think is wrong.
For example,the following lines are the output from 1(or 2) node calculation
and 6 nodes calculation. Look at the data. The atomic positions are the
same. But the order of atoms are different. Why dose this happen? Why the
results are different with 5 nodes and 6 nodes?
      atom type    atomic position
    1  Al    0.0000    0.0000   -1.8869
    2  O    -1.6539    0.1249   -1.0842
    3  O     0.7188   -1.4947   -1.0842
    4  O     0.9351    1.3699   -1.0842
    5  Al    2.3727   -1.3699   -0.2816
    6  Al   -2.3727    1.3699    0.2816
    7  O     1.6539   -0.1249    1.0842
    8  O    -0.7188    1.4947    1.0842
    9  O    -0.9351   -1.3699    1.0842
   10  Al    0.0000    0.0000    1.8869

    1  Al    0.0000    0.0000   -1.8869
    2  O    -1.6539    0.1249   -1.0842
    3  O     0.7188   -1.4947   -1.0842
    4  Al    0.9351    1.3699   -1.0842
    5  O     2.3727   -1.3699   -0.2816
    6  O    -2.3727    1.3699    0.2816
    7  O     1.6539   -0.1249    1.0842
    8  Al   -0.7188    1.4947    1.0842
    9  Al   -0.9351   -1.3699    1.0842
   10  Al    0.0000    0.0000    1.8869

Your help is really appreciated,
Wayne
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mpi-forum.org/pipermail/mpi-forum/attachments/20080707/039de7d4/attachment.html>