[openib-general] GAMESS and MVAPICH

Troy Benjegerdes
Tue Apr 20 11:04:48 PDT 2004


On Mon, Mar 29, 2004 at 05:45:29PM -0500, Dhabaleswar Panda wrote:
> Troy, 
> 
> You must have received a 2-line patch (related to memory
> registration/deregistration in MVAPICH2) from one of my students a
> short while ago. You can try that and let me know whether the problem
> is solved or not.
> 
> Please note that the MVAPICH2 release (0.5a) is an `alpha' release
> based on MPICH2 0.94b. I do not think we have touched anything related
> to MPI communicators.
> 
> If the problem persists, you can let us know the exact details of the
> errors you are seeing and we will be happy to take a look at it.
> 
> I will also suggest you to try GAMESS with MVAPICH. Based on our
> earlier conversation, I think it was running successfully with
> MVAPICH. Please let me know the status.
> 

We have successfully run GAMESS and gotten the correct answer on
MVIAPICH and mellanox vapi.

GAMESS uses a compute process and a data-server process model, and this
causes problems on MPI implementations that use polling. We've seen this
on Myrinet and Dolphin-SCI based MPI's, and now also on infiniband.

Is there a tuning parameter in one of the openib candidate stacks that
allows tuning of polling vs interrupt driven operation? At the moment
the data server winds up polling the infiniband card and taking cpu from
the compute process.

-- 
--------------------------------------------------------------------------
Troy Benjegerdes                'da hozer'                hozer at hozed.org  

Somone asked my why I work on this free (http://www.fsf.org/philosophy/)
software stuff and not get a real job. Charles Shultz had the best answer:

"Why do musicians compose symphonies and poets write poems? They do it
because life wouldn't have any meaning for them if they didn't. That's why
I draw cartoons. It's my life." -- Charles Shultz

-- 
To unsubscribe send an email with subject unsubscribe to openib-general at openib.org.
Please contact moderator at openib.org for questions.




More information about the openib-general mailing list