[mpich-discuss] intercommunicator support in MPICH

[Jing Fu]

Hi Jim,

The inter-communicator support (MPI_Comm_split) on bluegene should be fine.

I was splitting compute_comm and io_comm for some CFD application and
it runs just fine on Intrepid for upto 128k procs at least.

Like Rob said, point-to-point performance doesn't seem to be very
different, although collective communication on subcomms does
suffer a bit. On 64k procs on Intrepid, MPI_Allreduce for
integer/double on subcomms is  roughly 10 times slower than on
MPI_COMM_WORLD.

-----------------------------------------------------------------------------------
Jim Edwards jedwards at ucar.edu
Tue Jul 20 13:03:13 CDT 2010


Hi Rob,

I am getting different errors depending on the platform I am running on -
the only one that is clearly an mpi problem is edinburgh  with mpich-1.2.7p1
- so I will request an update to mpich2.

I am working on debugging on jaguar - if you have an account there maybe you
could have a look?  /tmp/work/jedwards/testpio/all.asb04    There is a
problem where something is not being communicated or is communicated
incorrectly, i've yet to find the source.

Bluegene also has problems I've yet to look into - I suspect I just need to
dig deeper.