[gmx-users] -rcon parameter in mdrun - what values have sense?

[Justin A. Lemkul]

Иимяа Фаамиилиияа wrote:
> Hi,
> 
> I constructed some (about 200 atoms) molecule, have got itp and gro files for it using PRODRG 
> server and put it in water box.

Is the topology just the raw output from PRODRG?  The parameters are often of 
poor quality and lead to unreliable simulations.  See, for instance:

http://pubs.acs.org/doi/abs/10.1021/ci100335w

I have occasionally observed complete crashes when using PRODRG parameters.

> Short test run was O'K. But when I tried long run I have got message:
> after about 1.5ns time: 
> ----------------------------------------------------------------------------------------------------------
> DD cell 0 1 0: Neighboring cells do not have atoms: 73
> DD cell 0 3 0: Neighboring cells do not have atoms: 50 49
> 
> Program mdrun, VERSION 4.5.5
> Source code file: domdec_con.c, line: 693
> 
> Fatal error:
> DD cell 0 1 0 could only obtain 43 of the 44 atoms that are connected via constraints 
> from the neighboring cells. This probably means your constraint lengths are too long 
> compared to the domain decomposition cell size. Decrease the number of domain 
> decomposition grid cells or lincs-order or use the -rcon option of mdrun. For more 
> information and tips for troubleshooting, please check the GROMACS
> website at http://www.gromacs.org/Documentation/Errors
> ----------------------------------------------------------------------------------------------------------- 
> I do not know where I can 
> 1) "...decrease the number of domain 
> decomposition grid cells or 

Use fewer threads/processors.

> 2) decrease lincs-order".
> 

This is specified in the .mdp file, but is likely not necessary.

> 3) I can use  -rcon option of mdrun.
> In this case I understand how I can change this parameter but I don't know what 
> are the reasonable values for it.
> Is it 0.761A  at the beginning of my calculation (see below)? 
> Should I try to make it shorter or longer?
> It looks that from one hand I should make it longer but from another hand it should be less
> than minimum cell size (0.898 in my case? see below)  
> 

I doubt you should be messing with these settings.  Your simulation appears to 
be crashing.  My first suspect is the topology from PRODRG, though in the 
absence of a complete .mdp file and description of the system, that's just 
conjecture at the moment.  See also:

http://www.gromacs.org/Documentation/Terminology/Blowing_Up

-Justin

> I have next information about domain decomposition in the beginning of run:
> ------------------------------------------------------------------------------------------------------------- 
> Initializing Domain Decomposition on 64 nodes
> Dynamic load balancing: auto
> Will sort the charge groups at every domain (re)decomposition
> Initial maximum inter charge-group distances:
>     two-body bonded interactions: 0.817 nm, LJ-14, atoms 138 148
>   multi-body bonded interactions: 0.817 nm, Proper Dih., atoms 148 138
> Minimum cell size due to bonded interactions: 0.898 nm
> Maximum distance for 5 constraints, at 120 deg. angles, all-trans: 0.761 nm
> Estimated maximum distance required for P-LINCS: 0.761 nm
> Using 32 separate PME nodes
> Scaling the initial minimum size with 1/0.8 (option -dds) = 1.25
> Optimizing the DD grid for 32 cells with a minimum initial size of 1.123 nm
> The maximum allowed number of cells is: X 5 Y 5 Z 5
> Domain decomposition grid 4 x 4 x 2, separate PME nodes 32
> PME domain decomposition: 4 x 8 x 1
> Interleaving PP and PME nodes
> This is a particle-particle only node
> -------------------------------------------------------------------------------------------------------------
> 
> 
> Any ideas what to do?
> Thank you in advance.
> Alex
> 
> 
> 
> 
> ----------------
> ----------------
> 

-- 
========================================

Justin A. Lemkul
Ph.D. Candidate
ICTAS Doctoral Scholar
MILES-IGERT Trainee
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin

========================================