[gmx-users] Transition difficulties: version 3.3.3 to 4.0.3 regarding pull_geometry=distance

Berk Hess gmx3 at hotmail.com
Sat Jan 24 18:58:15 CET 2009


Hi,

I just found out that I introduced a bug in 4.0.3, which could cause
the pull code to crash or give wrong results when running single processor.
In parallel it is correct.
I committed fixed for 4.0.4.

If you run in parallel things should be correct,
or change in src/kernel/md.c:
    if (ir->pull) {
      dd_make_local_pull_groups(NULL,ir->pull,mdatoms);
to
    if (ir->pull && PAR(cr)) {
      dd_make_local_pull_groups(NULL,ir->pull,mdatoms);

Berk

> Date: Fri, 23 Jan 2009 12:14:43 -0500
> From: fiedler at umich.edu
> To: gmx-users at gromacs.org
> Subject: Re: [gmx-users] Transition difficulties: version 3.3.3 to	4.0.3	regarding pull_geometry=distance
> 
> Hi,
> 
> Thank you Berk and Chris for the suggestions.
> 
> To address the possibility that this issue is related to periodic 
> boundaries, I used two approaches:
> 1.  The pull group of interest (permeant) was centered in the x-y plane 
> of the box using Chris' approach.  I then used the genconf utility to 
> replicate my lipid box to a 9x9 grid in the x-y plane and removed all 
> but the center box.  This generated the coordinates for a bilayer system 
> with all lipid molecules inside a box and intact.  The discrepancy 
> between the grompp (version 4.0.3) output and distances as calculated by 
> g_traj (version 4.0.3) persist, 2.667 vs. 0.3996 nm. 
> 2.  I constructed a three atom system containing 2 reference atoms of 
> type A, and a "pull" atom of type B.  Proper output from grompp was 
> observed for all coordinates of both the reference and pulled atoms, 
> include coordinates for atoms moved outside the box in the x-y plane.  
> The coordinate, topology, and run control parameter file are given below.
> 
> If there are additional suggestions, I would be greatly appreciative.
> 
> Thank you,
> 
>    Steve Fiedler
> 
> -----------------
> conf.gro
> Three atoms
>     3
>     1AAA      A    1   1.500   1.500   1.000
>     2AAA      A    2   0.500   1.500   1.000
>     3BBB      B    3  -1.500   1.500   1.700
>    3.00000   3.00000   3.00000
> -----------------
> index.ndx
> [ System ]
>    1    2    3
> [ Ref ]
>    1    2
> [ Pulled ]
>    3
> -----------------
> grompp.mdp
> title                    = ThreeAtoms
> integrator               = md
> dt                       = 0.001
> nsteps                   = 1
> ns_type                  = grid
> pbc                      = xyz
> coulombtype              = shift
> rlist                    = 1.4
> rcoulomb                 = 1.4
> rvdw                     = 1.4
> tcoupl                   = no
> pcoupl                   = no
> constraint_algorithm     = shake
> shake_tol                = 1e-4
> gen-vel                  = no
> gen-temp                 = 0
> 
> nstxout                  = 1
> nstvout                  = 0
> nstfout                  = 0
> 
> pull = umbrella
> pull_geometry = distance
> pull_dim = N N Y
> pull_start = no
> pull_init1 = 0.7
> pull_group0 = Ref
> pull_group1 = Pulled
> pull_k1 = 10000
> -----------------
> topology.top
> ; topology for two partially charged atoms
> 
> [ defaults ]
> ; nbfunc comb-rule gen-pairs fudgeLJ fudgeQQ
> 1     3     yes      0.125  0.5
> 
> [ atomtypes ]
> ;name     mass     charge ptype  sig           eps
>    A   1000.0000    0.000  A     0.50000      9.90000
>    B      9.0000    0.000  A     0.30000      9.00000
> 
> [ nonbond_params ]
>   ; i    j    func    sig          eps
> 
> [ moleculetype ]
> AAAA 1
> 
> [ atoms ]
> ;   nr   type  resnr residue  atom   cgnr  charge       mass
>     1     A     1      AAA    A      1      0.000   1000.0000
> 
> [ moleculetype ]
> BBBB 1
> 
> [ atoms ]
> ;   nr   type  resnr residue  atom   cgnr  charge     mass
>     1     B     1      BBB    B      1      0.000      9.00
> 
> [ system ]
> ; name
> Three atoms
> 
> [ molecules ]
> ; name   number
> AAAA      2
> BBBB      1
> 
> 
> 
> 
> 
> Chris Neale wrote:
> > I just checked similar simulations of mine and Berk's suggestion 
> > accounts for similar discrepancies that I notice on a quick evaluation 
> > where g_traj and g_dist fail to give me the same distance as I obtain 
> > from the pull pos.xvg file. As Berk suggests, once I first trjconv 
> > -center -pbc mol -ur compact (giving an appropriate residue for 
> > centering that puts all relevant pulled atoms in the same box) then 
> > g_traj and g_dist both give me the exact same answer as I calculate 
> > based on pull pos.xvg. Chris -- original message -- Hi, There could be 
> > a problem with periodic boundary conditions. Do you have multiple 
> > molecules in a pull group, or broken molecules? In that case the COM 
> > position of 3.3.3 and g_traj are both incorrect. The pull code in 4.0 
> > grompp and mdrun are (as far as I know) always correct. Berk
> >
> >> > Date: Thu, 22 Jan 2009 13:22:24 -0500
> >> > From: fiedler at umich.edu
> >> > To: gmx-users at gromacs.org
> >> > Subject: [gmx-users] Transition difficulties: version 3.3.3 to 
> >> 4.0.3    regarding pull_geometry=distance
> >> > > Dear all,
> >> > > I have encountered an odd behavior with use of the "pull_geometry 
> >> = > distance" option of the pull code, upon transitioning from 
> >> Gromacs > version 3.3.3 to version 4.0.3.  It appears to be related 
> >> to the center > of mass distances of the two pull groups, which has 
> >> an effect of > abruptly displacing the coordinates of the less 
> >> massive group.  A > diagnostic is a discrepancy between the distances 
> >> between the pull > groups from the preprocessor output in version 
> >> 4.0.3, and the distance > between the groups as calculated using the 
> >> difference of the groups' > centers of mass from the g_traj utility.  
> >> For example, using the > coordinates of a system previously 
> >> equilibrated with the constraint > force approach from version 3.3.3, 
> >> the grompp output from version 4.0.3 is:
> >> > > Pull group  natoms  pbc atom  distance at start     reference at t=0
> >> >        0      2672      1336 >        1        60      6818   
> >> 2.673                 0.400   > > Using g_traj (4.0.3 version), the 
> >> difference of the distance between the > center of masses of the two 
> >> groups is: 0.39911 nm versus the 2.673 value > from above. > > This 
> >> issue does not exist in previous versions of Gromacs including > 
> >> version 3.3.3.  In version 4.0.3, this behavior occurs for both > 
> >> pull=umbrella and pull = constraint, on 32 and 64 bit architecture > 
> >> systems, and in both single and double precision calculations.  A 
> >> test > of a two atom system determined that the pull_start option was 
> >> not > appropriate.  The pull options used in the mdp file are listed 
> >> below, as > well as the contents of the ppa file which has worked 
> >> previously. > > Suggestions would be appreciated,
> > _______________________________________________
> > gmx-users mailing list    gmx-users at gromacs.org
> > http://www.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
> >
> >
> 
> 
> -- 
> Steve Fiedler, Ph.D.
> Research Fellow
> Department of Mechanical Engineering
> The University of Michigan
> 2024 G.G. Brown
> 2350 Hayward St.
> Ann Arbor, MI 48109-2125
> 
> 
> _______________________________________________
> gmx-users mailing list    gmx-users at gromacs.org
> http://www.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

_________________________________________________________________
Express yourself instantly with MSN Messenger! Download today it's FREE!
http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20090124/7447a4d5/attachment.html>


More information about the gromacs.org_gmx-users mailing list