[gmx-developers] walls and pbc setting in CVS code

[Makoto Yoneya]

Hi,

The error message tells you (nearly) everything.
I guess you have a very small box and you have a high lincs_order and/or 
lincs_iter.
The error message tells you what to do.
Use -rcon
and/or decrease lincs_order and/or lincs_iter
or increase the box size.

Berk


> -----Original Message-----
> From: Makoto Yoneya [mailto:makoto-yoneya at aist.go.jp] 
> Sent: Monday, August 25, 2008 7:05 PM
> To: 'gmx-developers at gromacs.org'
> Subject: walls and pbc setting in CVS code
> 
> Dear GROMACS developers (Prof. Berk Hess):
> 
> Here again about the same subject as I posted in the last month.
> I'd tried the following modification  (shown by Prof. Hess) 
> of the CVS code.
> 
> > >In src/kernel/readir.c you have to comment out the line:
> > >  CHECK(ir->nwall && ir->ePBC!=epbcXY); and recompile grompp.
> 
> It works well and my original purpose (walls function with pbc=xyz.
> whereas, only pbc=xy was permitted with the falls function in 
> the original code) would be accomplished by this modification 
> (thanks a lot!).
> 
> In the course of the trial, I found strange behaviors and 
> I'll report for possible (?) bugfix.
> The code modification above was simply eliminates the 
> checking on the pbc for the walls function.
> However if I'd tried walls function with pbc=xy, my parallel 
> job (with 4
> nodes) was terminated with the following massage (this was 
> same in the original code without the modification).
> 
> --- begin the massage copy ---
> Initializing Domain Decomposition on 4 nodes Will not use 
> dynamic load balancing Will sort the charge groups at every 
> domain (re)decomposition Maximum distance for 9 constraints, 
> at 120 deg. angles, all-trans: 1.487 nm Estimated maximum 
> distance required for LINCS: 1.487 nm This distance will 
> limit the DD cell size, you can override this with -rcon 
> Optimizing the DD grid for 4 cells with a minimum initial 
> size of 1.487 nm The maximum allowed number of cells is: X 4 Y 4 Z 1
> 
> -------------------------------------------------------
> Program mdrun-mpi_d, VERSION 3.3.99_development_200800503 
> Source code file: domdec_setup.c, line: 372
>    
> Fatal error:
> There is no domain decomposition for 4 nodes that is 
> compatible with the given box and a minimum cell size of 
> 1.4875 nm Change the number of nodes or mdrun option -rcon or 
> your LINCS settings Look in the log file for details on the 
> domain decomposition ------end of the massage copy--------
> 
> The corresponding parallel job with simply changing pbc=xyz 
> runs normally without any errors.
> With pbc=xy, single node job made no errors, then possibly 
> the problem would be parallel decomposition with pbc=xy setting.
> Hope it helps.
> 
> Thanks again the precious efforts of GROMACS developers.
> 
> Makoto Yoneya, Dr.
> AIST, Japan
> 
> > -----Original Message-----
> > From: Makoto Yoneya [mailto:makoto-yoneya at aist.go.jp]
> > Sent: Tuesday, July 22, 2008 11:17 AM
> > To: 'gmx-developers at gromacs.org'
> > Subject: walls and pbc setting in CVS code
> > 
> > Dear Prof. Berk Hess and GROMACS developers:
> > 
> > Thanks a lot for your reply and the precious info.
> > Sorry for my bad explanations.
> > I simply want to simulate the system with a water layer 
> (modeled by a 
> > wall) and a monolayer on it.
> > So far the system can simulate with original CVS code.
> > However, I also want to add trans-monolayer constant vapor flow (in 
> > Z-direction) of water molecules from the water layer
> > (wall) to the system above.
> > I'll try the code modifications and check the result.
> > 
> > Best regards,
> > 
> > Makoto Yoneya, Dr.
> > AIST Japan
> > 
> > > Hello,
> > >
> > >I don't understand exactly what you want to do.
> > >But if you are really convinced that it makes sense for your
> > research,
> > >you can modify the code.
> > >
> > >In src/kernel/readir.c you have to comment out the line:
> > >  CHECK(ir->nwall && ir->ePBC!=epbcXY); and recompile grompp.
> > >
> > >Berk.
> > >
> > >
> > >Makoto Yoneya wrote:
> > > Dear GROMACS developers:
> > >
> > > I'd tried the walls function in the CVS code and found works well.
> > > In the current implementation, the walls function can 
> only be used 
> > > with pbc=xy as in the online manual.
> > > However, I'd like to use the walls without the pbc=xy 
> setting (i.e.
> > > the common 3-D pbc).
> > > I'm considering a Langmuir monolayer (monolayer on a water
> > layer) and
> > > trans monolayer water (in Z-direction from the water
> > > layer) simulation.
> > > The water layer may be modeled with the wall function 
> with the 9-3 
> > > potential.
> > > The trans monolayer water vapor flow (in Z-direction) may
> > be modeled
> > > with specifying,
> > > energygrp_excl: VAPOR_WATER wall0
> > > and
> > > acc_grps: VAPOR_WATER
> > > accelerate: 0.0 0.0 0.5.
> > > So, the water molecules may flow periodically in 
> Z-direction (then 
> > > requires 3-D pbc).
> > >
> > > If these may work, then please consider to work the walls 
> function 
> > > even without pbc=xy.
> > >
> > > Thank you for advance.
> > >
> > > Makoto Yoneya, Dr.
> > > AIST, Japan
> > >
> > >
> > > _______________________________________________
> > > gmx-developers mailing list
> > > gmx-developers at gromacs.org
> > > http://www.gromacs.org/mailman/listinfo/gmx-developers
> > > Please don't post (un)subscribe requests to the list. Use the www 
> > > interface or send it to gmx-developers-request at gromacs.org.
> > >
> > > This email was Anti Virus checked by Astaro Security Gateway. 
> > > http://www.astaro.com
> > >
> > >   
> > 
>