[gmx-users] Simulation runs on iMac but explodes on cluster

[Luke Goodsell]

Hi Mark,

Thanks for the information and help.

OK, but that can be just lucky - how lucky depends on your preparation
> protocol, which you haven't told us about. You are not actually doing the
> same calculation on both machines - the processes are chaotic even if
> everything seems equivalent.
>

The tpr file were provided by a collaborator who uses only Mac/OS X and so
has no experience with this problem. I don't know the preparation protocol,
but can ask if it is necessary.

Those warnings look pretty localized to a region of your structure. Firstly,
> some kind of linking or GROMACS failure would be likely to be manifest all
> over the place - and cause problems even if you did a run with different
> velocity generation. Secondly, have you looked at the resulting PDB files?
> You may wish to run the simulation to step 14500, and then restart with
> nstxout=1 so you can watch it explode. That may highlight a bad contact
> suddenly being able to relax.
>

Ahh, ok. The trajectory output is only every 500 steps, and the structures
up to step 14500 appear sensible.

Per's suggestion of disabling kernel optimisations appears to have caused
the simulation to run correctly so far. If this fails, I will follow your
suggestions to get more detailed information.

Again, many thanks for the advice.

Luke
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20110713/2c85533c/attachment.html>