[gmx-users] domain decomposition error
Justin A. Lemkul
jalemkul at vt.edu
Thu Feb 17 17:21:04 CET 2011
> Hello Justin,
> I am running in parallel but this error comes up with 8, 7, 6, 5, 4 and
> even 3 number of processors. I read about domain decomposition on the
> website and tried less number of processors, as just mentioned. My
> understanting after reading about the error is that parallel simulation
> is not working with a small system as mine! but there are 11000 atoms in
> there and I am doing FE stuff... I dont see why this is considered as
> small simulation!
This exact situation was just discussed last week regarding FEP and DD (hint:
check the archive!). If you set your intramolecular interactions to be computed
as exclusions and pairs (i.e. couple_intramol=no), then you have (essentially)
long-range bonded interactions. Either use mdrun -pd or run in serial.
> There is no domain decomposition for 4 nodes that is compatible with the
> given box and a minimum cell size of 5.70044 nm
> Change the number of nodes or mdrun option -rdd or -dds
> Also I checked the manual for the options below but I have no clue what
> value to set for my system.
> -rdd real 0 The maximum distance for bonded interactions
> DD (nm), 0 is determine from initial coordinates
> -dds real 0.8 Minimum allowed dlb scaling of the DD cell size
> BTW: About mdp setting for electrostatics, I have set zero charges to
> all atoms so there is no electrostatics term ( I have only C and H).
> Thats why I commented out ;coulombtype = PME
For both solute and solvent? Just one? Seems like an unreasonable model to me.
Even hydrocarbons should have some (obviously offsetting) partial charge.
Justin A. Lemkul
ICTAS Doctoral Scholar
Department of Biochemistry
jalemkul[at]vt.edu | (540) 231-9080
More information about the gromacs.org_gmx-users