Serge <skhan_at_[hidden]> writes:
> This is exactly what I am doing -- controlling distribution of
> processes with mpirun on the SGE-allocated nodes, by supplying the
> hostfile. Grid Engine allocates nodes and generates a hostfile, which
> I then can modify however I want to, before running the mpirun
> command. Moreover, it gives more control, by allowing to create
> specific SGE parallel environments, where the process distribution is
> predetermined -- one less worry for users playing with mpirun options.
I'm not sure I understand, but if you want control just via the PE
without using the allocation_rule (and without having to change the
mpirun invocation): As far as I remember from when it was done like
that here, if you use the PE starter to mangle SGE's $PE_HOSTFILE, which
the tight integration picks up, open-mpi 1.3 does the expected thing
with it, as for 1.2. Is that actually not the case, or is it just doing
it explicitly on the command line that's the problem?