why mpiexec run uncorrectly on smp?

[Pete Wyckoff]

ben at bellatrix.pcl.ox.ac.uk said:
> On Tue, Jun 11, 2002 at 03:33:55PM -0700, Brooks Davis wrote:
> > I'm pretty sure it's not an mpich limitation because mpirun works
> > fine on one machine without comm=shared.
> 
> 	Ah, but as I understand it, mpiexec uses MPICH's execer
> interface to start the processes, and mpirun doesn't. I guess the
> limitation is in there somewhere.

The problem is, I think, that mpiexec assumes that if you've got SMP
nodes, you're using mpich comm=shared, which perhaps isn't such a good
idea.  Perhaps it wouldn't be too hard to add a configure time option
to turn this off.  I don't think there's any run-time way to discern
the capabilities of libmpich.a, however.

I think it's just a matter of getting rid of the call to cull_nodes() in
tasks_shmem_reduce(), but I'm probably missing something...

		-- Pete