[gmx-users] Gromos DPPC bilayer: different results for 4.0.7 and 4.6.5

[Mark Abraham]

On Wed, Dec 18, 2013 at 6:58 AM, Lutz Maibaum <lutz.maibaum at gmail.com>wrote:

> Thank you all for your comments! It's interesting that Patrick ran into
> the same issues just a few days earlier.
>
> Following your suggestions, I started some simulations with the
> "Reaction-Field-nec" coulombtype. So far those have run for only about 30
> nanoseconds, but based on the attached graph I would assume that they will
> still result in an area per lipid that is lower than that obtained with
> Gromacs 4.0.7.  I will update the bug report on redmine once the
> simulations have progressed a little more.
>
> What would the next step be trying to debug this issue?


I would suggest using mdrun -rerun magic-configuration.gro -s some.tpr on
whatever combinations of .mdp settings and GROMACS versions make sense, to
observe that there is a difference that can be reproduced, and which .edr
fields show it. Then I'd suggest browsing the release notes for any obvious
clues, once there's a more clear idea what to look for (e.g. 4.0.7 was
fortuitously correct despite having a bug that was fixed). In extremis, one
could use a git bisect to walk back through history and find the commit in
which a difference arose, but doing that is reasonable only when one can
write a script that executes reasonably fast that can serve to identify
whether the behaviour is old or new (hence the mdrun -rerun suggestion;
equilibrating a membrane is not on!). Once the code change is identified,
the correctness of one version the code can be assessed.

Is it possible to output only the various components of the forces (in
> particular, those induced by the reaction field)?
>

Not directly. The non-bonded kernels for RF compute both 1/r and the
correction terms before recording any data. Note that the RF terms need not
be the primary problem - using the wrong form of long-range correction
would be a possible candidate, and there's doubtless others.

Mark


>
> Best,
>
>   Lutz
>
>
> --
> Gromacs Users mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
> posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
> send a mail to gmx-users-request at gromacs.org.
>