[gmx-users] dynamic selection memory allocation problem

[Teemu Murtola]

Please file a report a <http://redmine.gromacs.org> with files and the
exact selection necessary to reproduce the problem (a single .gro structure
is likely sufficient). While your approach is likely a very inefficient way
to use the selections (plus/merge is not designed for massive copying, and
your selection probably evaluates some parts multiple times over all the
atoms in the system), it would be useful to find out what is the actual
problem.

Your example is not particularly illustrative, as there is no point in
computing the same distance 15 times...
On Jun 27, 2014 3:34 PM, "gigo" <gigo at ibb.waw.pl> wrote:

> Dear GMXers,
> I noticed, that the dynamic selection method, for example in gmx distance,
> easily fails because of the memory allocation problem.
> I wanted to call gmx distance from a python script, which builds really
> long selections. An illustrating example, built of the same pairs of atoms
> connected by 'plus':
>
> gmx distance -s eq03.tpr -f eq03.xtc -oall qweqwe.xvg -xvg none -select
> '((resid 14497 and resname SOL and name LP1) or (resname LIG and name P))
> plus ((resid 14497 and resname SOL and name LP1) or (resname LIG and name
> P))'
>
> This one works. However, if you add several '((resid 14497 and resname SOL
> and name LP1) or (resname LIG and name P))' blocks more (13 in this
> particular case), soon you will see something like:
>
> *** Error in `gmx': realloc(): invalid next size: 0x0000000000fa57e0 ***
> Abort
>
> The rationale behind using the dynamic selection was simplification of the
> script that I've been working on - no need for writing and using an index
> file. I want to calculate about 40 different distances, which is currently
> far beyond the reach of gmx distance. I'm using gromacs-5.0-rc1. All limits
> related to memory allocation (stacksize etc) were removed. I tried it on a
> computer with loads of memory (and the number shown in the error message is
> rather small).
> If you have any suggestions, I would be very grateful.
> Best Regards,
>
> Grzegorz Wieczorek
> --
> 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.
>