Re: [Condor-users] Condor & MPICH2

["Erik Paulson" <epaulson@xxxxxxxxxxxx>]

On 2/14/07, Natarajan, Senthil <senthil@xxxxxxxx> wrote:
> Thanks Eric.
>
>
>
> So once if the mp1script (MPICH2 launcher) is fixed to run MPICH2. In order to submit parallel jobs, do we need to use
>
> executable = mp1script
>
>
>
> something like this in the job submission file?

That's mostly correct, though it is likely that the MPICH2 launcher
script would be named mp2script.

You'll also probably need to have something like
arguments = my_mpich2_linked_executable arg1 arg2
should_transfer_files = yes
when_to_transfer_output = on_exit
transfer_input_files = my_mpich2_linked_executable

Remember, the parallel universe in Condor just finds N machines, and
then launches a program on all N of them at the same time. If those
programs need to communicate, it's the job of the wrapper script to
setup anything that is needed. Condor includes some tools to make it
easier for the wrapper script to do that, and provides wrapper scripts
for LAM and MPICH1, but that's it.

Just to be clear, as far as I know, no one at the UW is currently
working on an MPICH2 wrapper script. Hopefully someone has one out
there that they can share.

-Erik