[mpich-discuss] MPICH2 works if executed on single machine buthangs over network.

[John Tomko]

Angus,

Which channel are you using? 
We encountered similar problems when using -channel shm .
We resolved the problem by lowering the priority in our program, so that
other processes can run, and every thing seemed to work.

John



-----Original Message-----
From: mpich-discuss-bounces at mcs.anl.gov
[mailto:mpich-discuss-bounces at mcs.anl.gov] On Behalf Of Angus Grandison
Sent: Friday, August 21, 2009 12:34 PM
To: mpich-discuss at mcs.anl.gov
Subject: [mpich-discuss] MPICH2 works if executed on single machine
buthangs over network.

I have used MPICH2 on a number of windows "clusters" with no problem.

I have set up a new cluster and it seems to work if the program has 2 
processes on a single computer but fails if I use 2 computers. I have 
debugged my own code and found that it hangs on the first MPI_BCast it 
encounters.  The cpi program appears to behave in a similar fashion e.g 
After entering an interval the program hangs even if the interval is
zero.

I am using MPICH2 (32bit) on a set Windows XP 64 machines. Just to note 
that this setup works successfully on a different set of Windows XP 64 
machines.

Are there any known problems with network cards and how they need to be 
setup?  The cards in my problem machines are INTEL Pro/1000 EB with I/O 
acceleration.

Can I use MPICH2 (64 bit) with 32bit executables?

TIA

Angus

-- 
_______________________________________________________________________

Dr Angus Grandison
Fire Safety Engineering Group
School of Computing and Mathematical Sciences
University of Greenwich
Old Royal Naval College
30 Park Row
Greenwich SE10 9LS
London
UK

web (group)    :   http://fseg.gre.ac.uk
web (personal) :   http://staffweb.cms.gre.ac.uk/~a.j.grandison/

mailto:A.J.Grandison at gre.ac.uk

Phone : +44 (0)20-8331-7912
Fax :   +44 (0)20-8331-8925
_______________________________________________________________________


University of Greenwich, a charity and company limited by guarantee,
registered in England (reg no. 986729).  Registered Office: Old Royal
Naval College, Park Row, Greenwich SE10 9LS.