You probably want to use the rank file mapper and/or the slot_list capability.
See the following for more info:
http://www.open-mpi.org/faq/?category=tuning#using-paffinity-v1.3
On Feb 20, 2012, at 1:05 PM, Jeff Layton wrote:
> Good afternoon,
>
> I'm working with some AMD Interlagos nodes (1
Good afternoon,
I'm working with some AMD Interlagos nodes (16-core, 4-socket),
and I want to bind processes to cores in specific layouts. These
may not be in logical order so I need to be able to my MPI app
and point it to specific cores on specific nodes (although the
core ordering is the same
