[gmx-developers] VFtab index too large

[Chenyue Xing]

I actually tested my new interaction on a 3-atom system and it worked fine.
Then I moved on to this complicated system.
It looks like the system takes quite a bit steps to get out of control.
Do you think rsq=dx*dx+dy*dy+dz*dz and rsq=dz*dz will have a very big
difference? As r will be used as r^6 and r^12 (r^7 and r^13 for force), I
assume the <1 value difference will have even smaller influence on
the energy and force part.

I really appreciate your help and hopefully you could point out my mistake
somewhere.

Thanks,
Chenyue


On 9/23/06, David van der Spoel <spoel at xray.bmc.uu.se> wrote:
>
> Chenyue Xing wrote:
> > My neighborlist is set to be updated every 10 steps. It core dumped in
> > the 4th loop (outloop index n=4). From the printed-outs, in the first
> > 3 loops, the neighborlists seem to be stable.
>
> if your atoms fly away there is something wrong :)
> are you sure your forces are correct?
> you should test the algorithm on a simple system and check for energy
> conservation.
>
> >
> > On 9/23/06, *David van der Spoel* <spoel at xray.bmc.uu.se
> > <mailto:spoel at xray.bmc.uu.se>> wrote:
> >
> >     Chenyue Xing wrote:
> >     > Hi David, Erik and other developers,
> >     >
> >     > You might remember I posted several questions related to the C
> >     codes
> >     > (nb_kernelxyz) weeks ago.
> >     > I've been trying to make a z-dependent only LJ interaction.
> >     > Once I make rsq=dz*dz (instead of rsq=dx*dx+dy*dy+dz*dz), after a
> >     > certain number of loops and steps, the coordinates of one neighbor
> >     > atom get quite large (of the order of 10^5), making rsq even
> larger.
> >     > As r is then used to computer the index of VFtab, nnn, this
> >     leads to a
> >     > nnn of the order of 10^9 and of course, the program cannot reach
> >     the
> >     > memory address of VFtab[1000000000]. So it goes to a
> >     segmentation fault.
> >     >
> >     > I really don't understand why a small change in rsq (my system is
> a
> >     > 128 DPPC with water, total 4700 coarse-grained atoms, dx dy is
> >     usually
> >     > around 1 nm for each step as printed out) will result in a
> >     sudden bump
> >     > of pos[] (or rvec x[]) values. If you could shine some light on
> this
> >     > issue, it will be very helpful.
> >     >
> >     > I appreciate your responses!
> >     >
> >     > Chenyue Xing
> >     How about your neighborlist? Is that updated as it should?
> >
> >     >
> >
> ------------------------------------------------------------------------
> >     >
> >     > _______________________________________________
> >     > gmx-developers mailing list
> >     > gmx-developers at gromacs.org <mailto:gmx-developers at gromacs.org>
> >     > http://www.gromacs.org/mailman/listinfo/gmx-developers
> >     <http://www.gromacs.org/mailman/listinfo/gmx-developers>
> >     > Please don't post (un)subscribe requests to the list. Use the
> >     > www interface or send it to gmx-developers-request at gromacs.org
> >     <mailto:gmx-developers-request at gromacs.org>.
> >
> >     _______________________________________________
> >     gmx-developers mailing list
> >     gmx-developers at gromacs.org <mailto:gmx-developers at gromacs.org>
> >     http://www.gromacs.org/mailman/listinfo/gmx-developers
> >     <http://www.gromacs.org/mailman/listinfo/gmx-developers>
> >     Please don't post (un)subscribe requests to the list. Use the
> >     www interface or send it to gmx-developers-request at gromacs.org
> >     <mailto:gmx-developers-request at gromacs.org>.
> >
> >
> > ------------------------------------------------------------------------
> >
> > _______________________________________________
> > gmx-developers mailing list
> > gmx-developers at gromacs.org
> > http://www.gromacs.org/mailman/listinfo/gmx-developers
> > Please don't post (un)subscribe requests to the list. Use the
> > www interface or send it to gmx-developers-request at gromacs.org.
>
> _______________________________________________
> gmx-developers mailing list
> gmx-developers at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-developers
> Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-developers-request at gromacs.org.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-developers/attachments/20060923/42a9206a/attachment.html>