[mpich-discuss] Assertion failure in ch3_progress

[Dave Goodell]

On Jan 30, 2009, at 3:45 AM, Dorian Krause wrote:

> b) It only occurs when the procs are on different machines.
>
> It would be helpful if there is a way to make sure that MPICH2
> behaves in the same way on shared memory and distributed memory
> machines (e.g. doesn't use IPC). Is there such a way (I suspect
> that there is a different behaviour because of point b))?

This can be controlled via an environment variable.  Setting  
MPICH_NO_LOCAL=1 will cause the network to be used for all  
communication and will disable any SMP-aware collective  
optimizations.  Obviously you will likely see a serious drop in  
performance.

You can also use MPICH_ODD_EVEN_CLIQUES=1 to make all odd processes  
appear to be on separate nodes while leaving the even processes on  
their current node.  So 4 processes on a node would be split into a  
total of 3 logical nodes as far as mpich2 is concerned.  One node  
would hold ranks 0 and 2, another node would hold rank 1, and a third  
node would hold rank 3.

-Dave