Re: FLUKA in MPI Environment

Hi Jing,

I am neither FLUKA nor MPI guru, but I can give you some advices:

In order to run a job at a MPI cluster, it has to be written for this
particular interface. You can not submit a FLUKA task at several CPUs
at the same time, but what you can do is to run several tasks in
parallel mode modifying the initial seed in the RANDOMIZe card. So,
all you need to do is to prepare several .inp files with different
initial seeds and submit them at your cluster separately.

There is another possibility which I find very convenient: one can
write a wrapper for the MPI interface which will distribute your FLUKA
tasks at different CPUs. In this case you need to submit just this
wrapper. Attached is a source code of the wrapper I had been using.

Actually, according to my experience, the MPI interface is not
convenient (and not designed) for running FLUKA tasks due to many
reasons. I think it's much more easier to use either GRID or LSF
platforms.

Konstantin