[gmx-users] Re: Re: nodes error

Peter C. Lai pcl at uab.edu
Fri Jan 6 10:39:56 CET 2012

By definition it is not really fixable.
You could lower some of the parameters that affects the minimum cell size
issue: lower your cutoffs and/or chargegroups. However this could affect
the validity of the resulting system.

Basically the problem here is that your cutoffs or chargegroup size is larger
than the minimum cell size generated by splitting your box across 192 
PP nodes and 64 PME nodes.

You could always add more molecules to increase the box dimensions, but
that will increase the amount of computatoin you have to do.

At some point you will run into diminshing returns anyway (i.e. your
efficiency will so bad you'll be essentialy wasting CPU cycles, not the
least because the nodes will spend a large amount of time doing 
synchronization). So if you can run the simulation with 128 CPU then
stick with that.

Finally, I don't necessarily trust the automated PME measurement that mdrun
computes. I've always manually allocated -npme based on what grompp told
me was an estimation of the PME:PP ratio based on the degrees of freedom
in the system (and I also manually set the domain decomposition too) but with 
bluegene you might not be able to allocate anything other than multiples of
32 nodes...

On 2012-01-06 10:23:26AM +0100, Albert wrote:
> is there any solution to fix this?
> On 01/06/2012 09:52 AM, gmx-users-request at gromacs.org wrote:
> > "The minimum cell size is controlled by the size of the largest charge
> > group or bonded interaction and the largest of rvdw, rlist and rcoulomb,
> > some other effects of bond constraints, and a safety margin. *Thus it is
> > not possible to run a small simulation with large numbers of processors.*"
> >
> > Based on the information you provided, the only thing I can say is, MAYBE
> > your system is "too small" to run with 256 processors.
> >
> > Cheers
> >
> > Terry
> -- 
> gmx-users mailing list    gmx-users at gromacs.org
> http://lists.gromacs.org/mailman/listinfo/gmx-users
> Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
> Please don't post (un)subscribe requests to the list. Use the 
> www interface or send it to gmx-users-request at gromacs.org.
> Can't post? Read http://www.gromacs.org/Support/Mailing_Lists

Peter C. Lai			| University of Alabama-Birmingham
Programmer/Analyst		| KAUL 752A
Genetics, Div. of Research	| 705 South 20th Street
pcl at uab.edu			| Birmingham AL 35294-4461
(205) 690-0808			|

More information about the gromacs.org_gmx-users mailing list