[gmx-users] gromacs 4.0.2 error in continuation to triclinic skew error

Berk Hess gmx3 at hotmail.com
Wed Nov 19 15:13:20 CET 2008


I don't know what the reason for the large LJ14 distance is,
but this will be over 3 bonds, not over 1 bond (which I guess the 0.154 is
that you mentioned).

This has nothing to do with the parallelization.
I would use 4.0.2, since there are many small issues fixed that were in 4.0.0.
In 4.0.0 you communicate twice as many atoms, so 4.0.2 will run faster.
I would suggest to run mdrun -rdd 1 and see if you don't get the missing
interaction errors anymore.


Date: Wed, 19 Nov 2008 14:06:17 +0000
From: lastexile7gr at yahoo.de
Subject: RE: [gmx-users] gromacs 4.0.2 error in continuation to triclinic skew	error
To: gmx-users at gromacs.org


thank you for all your responses up to now. The thing is getting much more clearer.

Yet, as far as the distance is concerned, this is a little bit worrying, because I have bonded interactions no more than 0.154 nm away. So a distance more than a whole nm I think that should be very long so as to keep this message to go away.  If this has to do with the parallelization, then  I might not worry. Yet, physicaly I think that is a very long distance.

I have to say that the same system worked very nicely with the version 4.0.0 and looking it via ngmx I saw nothing worrying. The properties of my system where reasonable too.

I insist, because in previous gromacs versions I got similar errors and their cause was due to running on many processors. The run was crashing, yet running it only in two
 processors I didn't face such problems. It is also curious because in version 4.0.0 I didn't see such an error. 

The runs I make take a lot of time, so seeing different error messages at random makes the post process even more difficult. The 1nm you propose I think that it is a big distance both for the LJ pairs interactions and of course for the bonded ones. Also I thought that the table extention was a cut-off for the LJ pairs interactions, which in my case is 5 nm.

The only thing I can think of for now is to run in less processors...

Thank you, 

Express yourself instantly with MSN Messenger! Download today it's FREE!
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20081119/07a92368/attachment.html>

More information about the gromacs.org_gmx-users mailing list