[gmx-users] mdrun -rerun

[Mark Abraham]

On 13/04/2012 10:44 AM, Juliette N. wrote:
> Hello all,
>
> I am trying to exclude a nonbonded interactions on the polymer chains using
>
> grompp -f old.mdp -c old_em.gro -p nrexcl_new.top -o new.tpr
>
> and mdrun -rerun command. but when I issue the command above grompp
> takes many hours to finish and at the end grompp crashes (Killed).
> This even leads to some weired hardware problem which gibes failure of
> node having some CPUs..I mean when I get grompp failure, I am not able
> to connect to the node anymore and will have to restart the node!
>
> grompp gives:
>
> Generated 332520 of the 332520 1-4 parameter combinations
> Opening library file /usr/local/gromacs/share/gromacs/top/spc.itp
> Opening library file /usr/local/gromacs/share/gromacs/top/ions.itp
> Excluding 101 bonded neighbours molecule type 'Polymer'
>
> Can anyone help me please?

Justin explained the fundamental reason for the problem. Apparently the 
implementation of the nrexcl=large is not robust with respect to memory 
usage and/or CPU time for large numbers of exclusions. So you will have 
to use energy group exclusions like I suggested earlier in this thread.

Mark

>
> Thanks
>
> On 8 April 2012 17:16, Justin A. Lemkul<jalemkul at vt.edu>  wrote:
>>
>> Juliette N. wrote:
>>> On 1 April 2012 20:17, Mark Abraham<Mark.Abraham at anu.edu.au>  wrote:
>>>> On 2/04/2012 10:10 AM, Juliette N. wrote:
>>>>> Hi all,
>>>>>
>>>>> I have an enquiry regarding calculation of heat of vaporization by
>>>>> estimating intermolecular nonbonded energies using mdrun rerun option.
>>>>> mdrun
>>>>> -rerun should break the total nonbonded energy coming from nonboded
>>>>> energy
>>>>> of (different molecules + a molecule with itself). By setting
>>>>> appropriate
>>>>> nrexcl in top file I am trying to exclude nonbonded part of molecule
>>>>> with
>>>>> itself within cut off radius so what remains would be intermolecular
>>>>> nonbonded energy between different molecules which determines heat of
>>>>> vaporization.
>>>>>
>>>>> 1) Is this approach correct?
>>>>
>>>> For excluding a whole molecule, it could work. For excluding only a part,
>>>> using energy group exclusions (see manual) is more flexible. Just setting
>>>> energy groups suitably might work in your case, so that you get the
>>>> group-wise break-down of nonbonded energy.
>>>>
>>>>
>>>>> 2) If yes, can you please check the way I am applying mdrun rerun:
>>>>>
>>>>> grompp -p nrexcl_3.top -o total_nonbonded.tpr
>>>
>>> Hello all,
>>>
>>> I am a bit confused about whether or not mdp file has to be provided
>>> for grompp rerun step. In the original run (no mdrun I provide mdp as
>>> follows:
>>>
>>>             grompp -f old.mdp -c  old_em.gro -p nrexcl_3.top -o
>>> total_nonbonded.tpr             (GROMPP old)
>>> then :
>>>             mdrun -deffnm total_nonbonded -s -o -c -g -e
>>>
>>> Then I update the top file to new nrexcl= new value
>>>
>>> No do I have to provide the old mdp file and old gro file old_em.gro,
>>> which were used in  (GROMPP old)? that is:
>>>
>>>   (GROMPP new:)       grompp -f old.mdp -c old_em.gro -p nrexcl_new.top
>>> -o new.tpr
>>>
>>> mpirun –np 4 mdrun_mpi -rerun total_nonbonded.trr  –deffnm new -s -o
>>> -c -g -e -x
>>>
>>>
>>>
>>> or I just have to use:
>>>   grompp   -p nrexcl_new.top -o new.tpr     ( no -c and no -f flag)
>>>
>> You need to provide some .mdp file and configuration.  If you omit -c and
>> -f, grompp (like any other Gromacs program) will search for the default
>> names of grompp.mdp and conf.gro.  If they don't exist, grompp will fail.
>>
>>
>>> My other question is that the grompp with large nrexcl around 100 is
>>> taking a lot of time, while the default  nrexcl=3 was grompp ed much
>>> faster. Why excluding bonds in this way is time consuming?
>>>
>> It's going slower because it's doing exponentially more work.
>>
>> -Justin
>>
>> --
>> ========================================
>>
>> 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
>>
>> ========================================
>>
>> --
>> 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/Support/Mailing_Lists/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/Support/Mailing_Lists
>
>