[gmx-users] gmx4.0.3: simulations do not run on 16 processors but do run on 10

[Mark Abraham]

Ángel Piñeiro wrote:
> Dear all,
> I am trying to run some simulations on our institutional cluster using
> gmx4.0.3. My systems are membrane proteins in lipid bilayers although
> for the error I am getting I do not think this is important (this is
> just my impression because a couple of months ago I got something
> similar with a completely different system). My simulation setups seem
> to be OK since I ran them locally (on one processor) with no problems
> (exactly the same tpr created in the cluster and copied to my PC) but
> when I submit the jobs to the queue of our cluster some simulations stop
> just after starting. This incidence seems to be a bit random although I
> noticed that some simulations that do not run on 16 processors do run on
> 10. For some cases a core file is generated together with traj.* and
> energy.edr files with 0 bytes but in other cases the simulation just
> stop without producing any information. Our administrator is
> investigating if it is a problem with MPI or the gmx installation but I
> would like to know if there is some known bug in gmx4.0.3 that could be
> responsible for this.

You should start by updating to 4.0.4 and investigating with that.

> I found this message in the gmx users mailing list that perhaps could be
> related... although it seems a bit old:
> http://www.gromacs.org/pipermail/gmx-users/2003-August/006793.html

Too old to be relevant.

Mark