[Pw_forum] Problem arising in fc_10.0.023+mkl_9.1.023

[Paolo Giannozzi]

On Oct 6, 2007, at 13:00 , Mahmoud Payami wrote:

> Dear PW users,
>
> I am testing the Intel fc_10.0.023 + mkl_9.1.023 on espresso_3.2.  
> [...]
> In running example_1, say, I encounter the message:
>
> ``running the symmetry analysis for Cu bands...[cli_0]: aborting job:
> application called MPI_Abort(MPI_COMM_WORLD,0) -process 1...''
> A crash file is also appeared containing the mewssage:
> ``from bands: error # 2
> pools not implemented''.
>
> However, the pools are checked and they work without any problem.

pools work without any problem where properly implemented, but
in some codes, pools are not implemented. More exactly: the
parallelization over k-points (requiring partition of processors into
the so-called pools) is not implemented. All codes call the same
initialization routine that partitions processes into pools if required
to do so with the appopriate command line option (-npool 2 for
instance), but unless somebody has programmed how to partition
the tasks and how to collect the results, pools will not work by
themselves. The code "bands.x" stops if run with pools because
unpredictable behavior may result. Apparently it doesn't stop in a
gracious manner, but it manages anyway to spread the message
why it is stopping.

Paolo
---
Paolo Giannozzi, Dept of Physics, University of Udine
via delle Scienze 208, 33100 Udine, Italy
Phone +39-0432-558216, fax +39-0432-558222