[gmx-users] Martini domain decomposition problem

[Szilárd Páll]

Hi Gábor,

Without more details (and little experience with martini systems) I can
suggest: tweaking DD so you don't end up with a too small box wrt the
number of cells by:
- Changing the number of ranks by using less/more cores to get a convenient
decomposition; I assume you had 6x4x4 before on 96 cores, so you can try
using e.g. 80 cores which will give you a 5x4x4 DD grid;
- Use OpenMP: your 6x4x4 DD with 1 OpenMP thread/rank will become 4x3x4
with 2 threads/rank.

Cheers,

--
Szilárd

On Wed, Sep 23, 2015 at 2:54 PM, Gábor Balogh <gaborbalogh91 at gmail.com>
wrote:

> Hello.
> I am trying to simulate a multi-domain protein using the Martini
> coarse-grained force field. The system consists of ~1600 protein "atoms"
> and ~26000 Martini water "beads". The water box is large compared to the
> size of the protein because of the significant conformational changes. The
> protein has an elastic network with the following parameters: lower cutoff:
> 0.5 nm, upper cutoff: 0.9 nm, force constant: 500 kJ/(mol*nm). Elastic
> "bonds" between atoms in two different domains were removed.
>
> The MD parameters are similar to this:
>
> http://cgmartini.nl/cgmartini/images/parameters/exampleMDP/martini_v2.x_new-rf.mdp
>
> My first simulation was running on a supercomputer, 96 CPU cores. The
> simulation crashed after ~2760000 steps:
>
> A list of missing interactions:
>        Harmonic Pot. of   2821 missing      1
>
> Molecule type 'Protein'
> the first 10 missing interactions, except for exclusions:
>        Harmonic Pot. atoms 1098 1306           global  1098  1306
> Fatal error:
> 1 of the 8124 bonded interactions could not be calculated because some
> atoms involved moved further apart than the multi-body cut-off distance
> (1.317 nm) or the two-body cut-off distance (1.317 nm), see option -rdd,
> for pairs and tabulated bonds also see option -ddcheck
>
> Second simulation (-rdd was set to 1.6, 72 CPU cores) crashed after
> ~70170000 steps:
>
> Fatal error:
> The X-size of the box (9.696262) times the triclinic skew factor (0.990125)
> is smaller than the number of DD cells (6) times the smallest allowed cell
> size (1.600000)
>
> Third simulation (rdd=1.5, 48 CPU cores) crashed after 24939840 steps:
>
> Step 24939840: The domain decomposition grid has shifted too much in the
> Y-direction around cell 2 3 0 (2 3 2, 3 3 0, 1 3 1, 2 3 1, 3 3 1, 1 3 2, 3
> 3 2, 1 3 0, 0 3 0, 0 3 2, 0 3 1). This should not have happened. Running
> with less nodes might avoid this issue. (12 error messages).
>
> What MD/domain decomposition parameters should I use?
>
> Thanks in advance.
> --
> Gromacs Users mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
> posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
> send a mail to gmx-users-request at gromacs.org.
>