[gmx-users] Re.Importance of -rcon and -dd options when

Mark Abraham Mark.Abraham at anu.edu.au
Thu Mar 15 08:04:09 CET 2012


On 15/03/2012 5:36 PM, PAVAN PAYGHAN wrote:
>
>
>     On 14/03/2012 9:54 PM, PAVAN PAYGHAN wrote:
>     >
>     > Dear Gromacs Users,
>     >
>     > I am running mdrun on single node with 8 CPU and getting
>     following error
>     >
>     > Fatal error:-
>     >
>     > D D cell  1 0 0 could only obtain 1520 of the 1521 atoms that are
>     > connected via constraints from the neighbouring cells.
>     >
>     > This probably means your constraint length are too long compared to
>     > the domain decomposition cell size.
>     >
>     > Decrease the number of domain decomposition grid cells or lincs
>     order.
>     >
>
>     The .log file has a detailed analysis of how DD is setting things up.
>     You need to make sure that this output is sensible for your system.
>
>     You should also desist from using P-R pressure coupling during
>     equilibration (i.e. with velocity generation), as warned in the manual
>     section on pressure coupling. Perhaps your system is blowing up.
>
>     Mark
>
>     Dear Mark,
>
> Thanks for the reply.
>  I have successfully done the equilibration run ,while doing 
> production run I am gettiing above error (almost after 10 ns run).
>  Also It would be worth working if you explain the importance -rcon 
> value and other things that I have asked .

A description of your system, GROMACS version and objective are 
important to solving the problem. I don't think we've seen those. Nobody 
wants to spend time talking about DD options if you are trying to run a 
system with 1000 atoms on 8 processors. By not describing fully, you 
make it easy for people to not be bothered.

>
> Regards,
>
> Pavan
>
>     > For solving this problem following are the attempts by me.
>     >
>     > *1] Decreasing grid cell size:-*
>     >
>     >  As per the suggestion in error, I tried to decrease the grid
>     cells by
>     > option -dd from 8 1 1 to 6 1 1 , it has thrown following error..
>     >
>     >  Fatal error:-The size of the DD grid (6) does not match the
>     number of
>     > nodes(8).
>     >
>     > Can you please suggest any better way to overcome this for
>     decreasing
>     > grid cell size.
>

You can't decrease the number of DD cells since you need one per 
processor. Maybe you are trying to parallelize a system that is too 
small for this number of processors, which brings us back to needing a 
description of your system.

>     >
>     > *2] -rcon option:*
>     >
>     > What is the correlation between the -rcon value with DD cell
>     > size(Directly or inversely proportional ) , for the problem entitled
>     > above what should be the strategy (to decrease or to increase
>     rcon value).
>     >
>     > If one changes the -rcon value will it affect the lincs accuracy, or
>     > in other words the run will hold the same continuation or any change
>     > in it.
>     >
>     > For changing the -rcon value the reference of previous log file i.e.
>     > Estimated maximum distance required for p-lincs say 0.877, so
>     one can
>     > increase than what has been estimated.
>

You need to read mdrun -h about -rcon. You need to be trying to increase 
the ratio of cell size to constraint length, per the error message.

>     >
>     > *3] lincs_order and lincs_iter :*
>     >
>     > **
>     >
>     > If we don't want to deteriorate the lincs accuracy (1+
>     > lincs_iter)*lincs_order has to remain constant , In my case
>     >
>     > With lincs_order = 4 and lincs_iter =1 I got above error. So I
>     > decreased lincs _order (2) and increased lincs_iter(3)
>     proportionally.
>     > What I am following is right or I have misunderstood it. If so
>     please
>     > correct it. Can this value be fraction?
>

The values must be integers.

>     > Values which I have tried are relevant or very bad?**
>

That is a correct approach for maintaining LINCS accuracy and trying to 
decrease the required constraint length, however it may not help solve 
the underlying problem.

Mark

>     >
>     > Please explain it.
>     >
>     > If the same problem can be solved by any other methodology please
>     > explain it.
>     >
>     > *Please see the mdp file  details.*
>     >
>     >
>     > integrator        = md
>     >
>     > nsteps             = 10000000
>     >
>     > dt                         = 0.002                   ; 2 fs
>     >
>     > ; Output control
>     >
>     > nstxout                        = 1000                        ; save
>     > coordinates every 2 ps
>     >
>     > nstvout                        = 1000                        ; save
>     > velocities every 2 ps
>     >
>     > nstxtcout         = 1000                        ; xtc compressed
>     > trajectory output every 2 ps
>     >
>     > nstenergy       = 1000                        ; save energies
>     every 2 ps
>     >
>     > nstlog              = 1000                        ; update log file
>     > every 2 ps
>     >
>     > ; Bond parameters
>     >
>     > continuation   = yes                   ; Restarting after NPT
>     >
>     > constraint_algorithm = lincs ; holonomic constraints
>     >
>     > constraints     = all-bonds             ; all bonds (even heavy
>     atom-H
>     > bonds)
>     >
>     > lincs_iter         = 1                               ; accuracy
>     of LINCS
>     >
>     > lincs_order     = 4
>     >
>     > ; Neighborsearching
>     >
>     > ns_type                       = grid
>     >
>     > nstlist              = 5
>     >
>     > rlist                  = 1.2
>     >
>     > rcoulomb        = 1.2
>     >
>     > rvdw                = 1.2
>     >
>     > ; Electrostatics
>     >
>     > coulombtype  = PME
>     >
>     > pme_order     = 4
>     >
>     > fourierspacing           = 0.16
>     >
>     > ; Temperature coupling is on
>     >
>     > tcoupl              = Nose-Hoover
>     >
>     > tc-grps                        = Protein  P    SOL_NA_CL ; three
>     > coupling groups - more accurate
>     >
>     > tau_t                = 0.5   0.5       0.5
>     >
>     > ref_t                = 323 323     323
>     >
>     > group, in K
>     >
>     > ; Pressure coupling is on
>     >
>     > pcoupl             = Parrinello-Rahman     ; Pressure coupling
>     on in NPT
>     >
>     > pcoupltype     = semiisotropic                         ; uniform
>     > scaling of x-y box vectors, independent z
>     >
>     > tau_p              = 2.0                                   ; time
>     > constant, in ps
>     >
>     > ref_p               = 1.0   1.0                           ;
>     reference
>     > pressure, x-y, z (in bar)
>     >
>     > compressibility = 4.5e-5       4.5e-5 ; isothermal
>     compressibility, bar^-1
>     >
>     > ; Periodic boundary conditions
>     >
>     > pbc                      = xyz                        ; 3-D PBC
>     >
>     > ; Dispersion correction
>     >
>     > DispCorr        = EnerPres    ; account for cut-off vdW scheme
>     >
>     > ; Velocity generation
>     >
>     > gen_vel                       = yes
>     >
>     > **
>     >
>     > Pavan
>     >
>     >
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20120315/4f8811bc/attachment.html>


More information about the gromacs.org_gmx-users mailing list