[mpich-discuss] a question about process-core binding

[Pavan Balaji]

On 08/02/2011 03:49 PM, teng ma wrote:
> Another problem is still -binding rr.  -binding rr looks still filling
> out 1st node's all  slots and then 2nd node's.  Its binding looks the
> same with cpu on my case.   Do I need some trick on hostfile to reach an
> effect of OpenMPI's -bynode or MVAPICH2's scatter?

What does the hostfile contain? Can you send it to us?

-- 
Pavan Balaji
http://www.mcs.anl.gov/~balaji