Much appreciated!

Per some of my other comments on this thread and on the referenced ticket, can you tell me what kernel you have on that machine? I assume you have NUMA support enabled, given that chipset?

Thanks!
Ralph

On Wed, Jun 10, 2009 at 10:29 AM, Sylvain Jeaugey <sylvain.jeaugey@bull.net> wrote:
Hum, very glad that padb works with Open MPI, I couldn't live without it. In my opinion, the best debug tool for parallel applications, and more importantly, the only one that scales.

About the issue, I couldn't reproduce it on my platform (tried 2 nodes with 2 to 8 processes each, nodes are twin 2.93 GHz Nehalem, IB is Mellanox QDR).

So my feeling about that is that is may be very hardware related. Especially if you use the hierarch component, some transactions will be done through RDMA on one side and read directly through shared memory on the other side, which can, depending on the hardware, produce very different timings and bugs. Did you try with a different collective component (i.e. not hierarch) ? Or with another interconnect ? [Yes, of course, if it is a race condition, we might well avoid the bug because timings will be different, but that's still information]

Perhaps all what I'm saying makes no sense or you already thought about this, anyway, if you want me to try different things, just let me know.

Sylvain


On Wed, 10 Jun 2009, Ralph Castain wrote:

Hi Ashley

Thanks! I would definitely be interested and will look at the tool. Meantime, I have filed a bunch of data on this in
ticket #1944, so perhaps you might take a glance at that and offer some thoughts?

https://svn.open-mpi.org/trac/ompi/ticket/1944

Will be back after I look at the tool.

Thanks again
Ralph


On Wed, Jun 10, 2009 at 8:51 AM, Ashley Pittman <ashley@pittman.co.uk> wrote:

     Ralph,

     If I may say this is exactly the type of problem the tool I have been
     working on recently aims to help with and I'd be happy to help you
     through it.

     Firstly I'd say of the three collectives you mention, MPI_Allgather,
     MPI_Reduce and MPI_Bcast one exhibit a many-to-many, one a many-to-one
     and the last a many-to-one communication pattern.  The scenario of a
     root process falling behind and getting swamped in comms is a plausible
     one for MPI_Reduce only but doesn't hold water with the other two.  You
     also don't mention if the loop is over a single collective or if you
     have loop calling a number of different collectives each iteration.

     padb, the tool I've been working on has the ability to look at parallel
     jobs and report on the state of collective comms and should help narrow
     you down on erroneous processes and those simply blocked waiting for
     comms.  I'd recommend using it to look at maybe four or five instances
     where the application has hung and look for any common features between
     them.

     Let me know if you are willing to try this route and I'll talk, the code
     is downloadable from http://padb.pittman.org.uk and if you want the full
     collective functionality you'll need to patch openmp with the patch from
     http://padb.pittman.org.uk/extensions.html

     Ashley,

     --

     Ashley Pittman, Bath, UK.

     Padb - A parallel job inspection tool for cluster computing
     http://padb.pittman.org.uk

_______________________________________________
devel mailing list
devel@open-mpi.org
http://www.open-mpi.org/mailman/listinfo.cgi/devel




_______________________________________________
devel mailing list
devel@open-mpi.org
http://www.open-mpi.org/mailman/listinfo.cgi/devel