[gmx-users] divergent energy minimization results from identical starting system
talcite at gmail.com
Sat Jan 29 07:15:01 CET 2011
Ah, the -reprod option makes everything consistent between runs now. The
term 'binary reproducibility' in the help message was a bit confusing at
first, but I found the wiki page on reproducibility.
Thanks for your help,
On 01/28/2011 06:29 PM, Roland Schulz wrote:
> are you running in parallel (either MPI or threads)? Load-balancing is
> one reason for different rounding errors.
> You can run with "mdrun -reprod" to avoid any different rounding
> between runs and should in general get the same then.
> On Fri, Jan 28, 2011 at 6:17 PM, Matthew Chan
> <mchan3 at connect.carleton.ca <mailto:mchan3 at connect.carleton.ca>> wrote:
> My second question is about the divergent energy minimization
> results which I have been receiving.
> I've taken the 1AKI lysozyme and prepared a single em.tpr file
> (1AKI is in vacuum). Afterwards I make 10 copies of the em.tpr
> file and use mdrun on each one. If I set the stopping condition to
> less than 1000kJ/mol nm, my potential energies and final
> structures from each run are not identical.
> I've tried both steepest descent and cg methods for minimization.
> I've also checked that the Fmax is indeed less than my stopping
> condition, and the potential energy is negative. Is this problem
> well documented or is there something wrong with my system? There
> seem to be a few parts of the manual that allude to the
> possibility of variance between subsequent runs of EM.
> If this is a well documented problem, can someone try explaining
> the cause to me please? I would like to learn more about this topic.
> Also, the potential energy value reported seems to be several
> orders of magnitude different from what other programs are
> reporting (24 000 vs 500). What units are it expressed in?
> Thanks in advance for your replies,
> Matthew Chan
> gmx-users mailing list gmx-users at gromacs.org
> <mailto:gmx-users at gromacs.org>
> 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
> <mailto:gmx-users-request at gromacs.org>.
> Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
> ORNL/UT Center for Molecular Biophysics cmb.ornl.gov <http://cmb.ornl.gov>
> 865-241-1537, ORNL PO BOX 2008 MS6309
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the gromacs.org_gmx-users