[gmx-users] Re: Issues regarding exclusions and Charge group distribution

Mark Abraham Mark.Abraham at anu.edu.au
Fri Jun 12 07:17:56 CEST 2009


Manik Mayur wrote:
> On Fri, Jun 12, 2009 at 8:46 AM, Mark Abraham <Mark.Abraham at anu.edu.au 
> <mailto:Mark.Abraham at anu.edu.au>> wrote:
> 
>     Manik Mayur wrote:
> 
>         Hi,
> 
>         The last mail supposedly bounced from the server due to
>         attachments. So please excuse in case you find this as a repetition.
> 
>         I am trying to simulate a system where I want to exclude all nb
>         interactions between frozen groups. So I followed the manual and
>         defined them in energygrp_excl. But upon doing that it throws
>         the following warning.
> 
>         WARNING 1 [file topol.top, line 1583]:
>          Can not exclude the lattice Coulomb energy between energy groups
> 
>         What is the reason of the warning and should I be worried about
>         it or safely ignore it.
> 
> 
>     So, what algorithm are you using for your long-ranged electrostatic
>     interactions? Do you know roughly how it works? Does it then make
>     sense to try to exclude some of the long-range interactions?
> 
> 
> I am using PME. I honestly do not have much idea about its 
> implementation details but I will definitely look into it. Here I was 
> trying to remove all interactions between the frozen group so that I can 
> speedup my simulation a little bit. Also since the atoms in the frozen 
> group are uncharged, my basic concern was to remove even the LJ 
> interaction calculation.

OK. The description in the GROMACS manual is OK on the maths and 
implementation details, but doesn't describe the underlying strategy. 
See http://en.wikipedia.org/wiki/Particle_mesh_Ewald for starters. The 
long-ranged component cannot be decomposed group-wise, so cannot be 
excluded group-wise. There are many posts in the archive on this topic.

> Till now I have ignored the warning. So is it ok?

Yes, but you're only saving time in your short-ranged electrostatic 
calculations.

>         Next, I am facing a strange problem with mdrun_mpi. If I run it
>         on 4 nodes then it throws up the following error. Which I traced
>         to md.log where I saw:
> 
>         There are: 15923 Atoms
>         Charge group distribution at step 0: 0 3264 3185 0 (why these
>         zeroes?)
> 
> 
>     You have atoms that are not part of a charge group (or, probably,
>     are part of one of size one and charge zero.
> 
> In my topology file, All the atoms that I have defined within a molecule 
> do belong to a charge group. Could you please elaborate on the second 
> part of your statement?

I'm not sure how the accounting works - GROMACS might not be counting 
charge groups that have only excluded interactions, or not counting 
charge groups that have no non-zero charges (and thus never participate 
in an electrostatic interaction). Whatever is going on, the workload 
looks like it might be unbalanced over the four processors.

Mark

>         Grid: 3 x 4 x 5 cells
> 
>         However with 2 nodes it has and runs fine:
> 
>         There are: 15923 Atoms
>         Charge group distribution at step 0: 3226 3223
>         Grid: 4 x 5 x 25 cells
> 
>         The error-
> 
> 
>     No, that's the dump from MPI after the error. You need to consult
>     both the stdout and the .log file.
> 
>     Mark
> 
>         Making 1D domain decomposition 1 x 1 x 4
>         starting mdrun 'TransverseElectrode'
>         100000 steps,    200.0 ps.
>         step 0
> 
>         NOTE: Turning on dynamic load balancing
> 
>         [md-comp1:32374] *** Process received signal ***
>         [md-comp1:32377] *** Process received signal ***
>         [md-comp1:32377] Signal: Segmentation fault (11)
>         [md-comp1:32377] Signal code: Address not mapped (1)
>         [md-comp1:32377] Failing at address: (nil)
>         [md-comp1:32374] Signal: Segmentation fault (11)
>         [md-comp1:32374] Signal code: Address not mapped (1)
>         [md-comp1:32374] Failing at address: (nil)
>         [md-comp1:32377] [ 0] [0xb7fd8440]
>         [md-comp1:32374] [ 0] [0xb7ef0440]
>         [md-comp1:32374] *** End of error message ***
>         [md-comp1:32377] *** End of error message ***
> 
>         I would be extremely thankful, if somebody points me to the
>         source of the error and the workaround.
> 
>         Thanks,
> 
>         Manik Mayur
>         Graduate student
>         Microfluidics Lab
>         Dept. of Mechanical Engg.
>         IIT Kharagpur
>         INDIA
> 
> 
>         ------------------------------------------------------------------------
> 
>         _______________________________________________
>         gmx-users mailing list    gmx-users at gromacs.org
>         <mailto:gmx-users at gromacs.org>
>         http://lists.gromacs.org/mailman/listinfo/gmx-users
>         Please search the archive at http://www.gromacs.org/search
>         before posting!
>         Please don't post (un)subscribe requests to the list. Use the
>         www interface or send it to gmx-users-request at gromacs.org
>         <mailto:gmx-users-request at gromacs.org>.
>         Can't post? Read http://www.gromacs.org/mailing_lists/users.php
> 
>     _______________________________________________
>     gmx-users mailing list    gmx-users at gromacs.org
>     <mailto:gmx-users at gromacs.org>
>     http://lists.gromacs.org/mailman/listinfo/gmx-users
>     Please search the archive at http://www.gromacs.org/search before
>     posting!
>     Please don't post (un)subscribe requests to the list. Use the www
>     interface or send it to gmx-users-request at gromacs.org
>     <mailto:gmx-users-request at gromacs.org>.
>     Can't post? Read http://www.gromacs.org/mailing_lists/users.php
> 
> 
> Manik Mayur
> Graduate student
> Microfluidics Lab
> Dept. of Mechanical Engg.
> IIT Kharagpur
> INDIA
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> gmx-users mailing list    gmx-users at gromacs.org
> http://lists.gromacs.org/mailman/listinfo/gmx-users
> Please search the archive at http://www.gromacs.org/search before posting!
> Please don't post (un)subscribe requests to the list. Use the 
> www interface or send it to gmx-users-request at gromacs.org.
> Can't post? Read http://www.gromacs.org/mailing_lists/users.php



More information about the gromacs.org_gmx-users mailing list