On Thu, Mar 01, 2012, David A Case wrote:
> ----- Forwarded message from Holger Gohlke <gohlke.uni-duesseldorf.de> -----
>
> > >
> > > 1. Many of the "decomposition" tests are failing. Is this because some
> > > radii were changed, or because some algorithm was changed? If there
> > > is anyone who uses this stuff and can tell what is going on, that would
> > > help a lot.
> >
> > We saw this as well when we wanted to check in final new .save files
> > for our examples. The reason must be within sander and was probably
> > introduced between 13.02.2012 and 28.02.2012. What is troublesome is
> > that the sum of the internal energies from the decomp. analysis is not
> > equal to the sum of these energies from sander.out anymore.
> >
I'd like to see if we can track this down.
1. James: can you check to see if changes you made to mdread.F90 in the last
month (log files refer to "GB parameter assignment") might have changed the
radii that are being used in the surface area calculation, (or in any other
way might be affecting ESURF)? I think you can just look at the tests in
$AMBERHOME/test/idecomp1 -- I guessing that if we find the problem there, it
will be fixed elsewhere as well. Since the get_atomic_number() code is all
new, it looks possible that the radii being assigned are not the same as they
were before the change; but the .save results in the decomp directories have
not been updated.
2. Holger: can you double check that the recent changes to decomp.F90 (commit
6ac8353095b0a, by N. Homeyer, did not do anything wrong in terms of how the
decomposed energies add up to the total energies?
3. Scott: can you poke through the Cruise Control logs to try to identify the
spot at which the decomposition tests cases started to fail? That should help
us pin down the code change that is causing the problem.
...thanks...dave
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Fri Mar 02 2012 - 06:00:04 PST
