[gmx-developers] Problem with PME in CVS

Justin MacCallum jlmaccal at ucalgary.ca
Tue Jul 5 21:57:58 CEST 2005


I've noticed a problem when running simulations using PME with the  
CVS version of gromacs. I can run simulations with 1,2,3, and 4  
processors, but when I try to run on 8 or 16 processors things aren't  
working correctly. Gromacs seems to start correctly, but hangs before  
the output line for step 0 is printed.  Mpich prints out the  
following error: "Timeout in establishing connection to remote  
process: 0". Strangely the logfile for nodes 0-3 end with:

Iniitalizing LINear Constraint Solver

While nodes 4-7 end with:

Started mdrun on node X DATE
Grid: ...
Configuring non-bonded kernels
Testing AMD 3DNow support... not present
Testing ia32 SSE support... present

So it appears that some of the nodes aren't initializing properly.  
This is run on a cluster of dual xeons running ROCKS. I should  
mention that all of these jobs run fine with 3.2.1, although the  
performance is obviously quite poor for more than two processors.


Justin MacCallum
PhD Student
Departement of Biological Sciences
University of Calgary

email: jlmaccal at ucalgary.ca
web: moose.bio.ucalgary.ca

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-developers/attachments/20050705/f2fc48ca/attachment.html>

More information about the gromacs.org_gmx-developers mailing list