[gmx-developers] problem with particle decompostion and with gromacs 4.6.3

[Mark Abraham]

On Thu, Oct 3, 2013 at 5:21 PM, Floris van Eerden <f.j.van.eerden at rug.nl>wrote:

> Dear reader,
>
> This post describes two different problems while I was trying to run a
> protein monomer.
>
> The first problem is that I get completely different results when I run a
> system with particle decomposition or domain decomposition.
> I simulate a coarse grained protein monomer embedded in a lipid bilayer
> and solvated in water+ions. When I use domain decomposition the system runs
> normal.
> However when I switch to particle dynamics the membrane starts buckling
> heavily, resulting in a completely folded membrane.
>

Other than the range of non-bonded interactions, is there anything
noteworthy in the topology or run setup? E.g. restraints, vsites, md-vv,
replica exchange, free-energy... The most likely problem is some feature
nobody ever tested / intended to work with PD. This is part of why PD is on
the chopping block for 5.0.


> This is in Gromacs 4.5.5 and 4.5.3
>

Please try a 4.5.7 run, so we can eliminate the years of accrued bug fixes
as the source of the problem :-)


> I tried to reproduce these errors in Gromacs 4.6.3, but here my
> simulations kept crashing. When i used particle dynamics I get a lot of
> lincs warnings followed by a segmentation fault and when i use domain
> decomposition it crashes with a 'software inconsistency error: some
> particles seem to be  assignment multipe times' (run on 8 cores with rdd
> set to 1.8). When I use only one thread (-n1 ) I get an segmentation fault.
> These while I can run these simulations without any problem (except the
> buckling) on gromacs 4.5.5. It does not matter if the tpr file is generated
> with gromacs 4.6.3 or gromacs 4.5.5.
>
> Has anybody an idea what  might be the problem?
>

Some more information about your simulation would be good, but we'd
probably like you to open an issue at redmine.gromacs.org so you can attach
.tpr + preparation files as a tarball there.

Thanks!

Mark
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-developers/attachments/20131003/29ec189e/attachment.html>