[gmx-users] Gromacs-4.0.2: energy minimization error in parallel version

[Suman Chakrabarty]

Hi Berk,

Thanks for the update.

I think steepest descent will be enough for me indeed. But I was not
really sure about the preferred value for emtol ... I was using 1.0
following a colleague! Thanks again.


Regards,
Suman.


On Sat, Nov 22, 2008 at 1:58 AM, Berk Hess <gmx3 at hotmail.com> wrote:
> Hi,
>
> I fixed the bug for version 4.0.3.
> There was also a bug with writing the forces at the end
> with nstfout=0.
>
> But if you are minimizing for starting a md simulation,
> you do not need cg, steepest descents is enough
> (with em_tol of 1000 or 100).
>
> Note that in 4.0 you can now minimize with constraints,
> box for water and other molecules.
>
> If you really need to run parallel cg now, you can use
> the -pd switch of mdrun.
>
> Berk
>
>
>
>> Date: Sat, 22 Nov 2008 00:37:18 +0530
>> From: suman at sscu.iisc.ernet.in
>> To: gmx-users at gromacs.org
>> Subject: [gmx-users] Gromacs-4.0.2: energy minimization error in parallel
>> version
>>
>> Dear all,
>>
>> I am trying to use mdrun_mpi for energy minimization over 2/4/8 nodes.
>> But in every case the run gets aborted with the following error:
>>
>> -------------------------------------------------------
>> Program mdrun_mpi, VERSION 4.0.2
>> Source code file: minimize.c, line: 404
>>
>> Software inconsistency error:
>> state mismatch in do_em_step
>> -------------------------------------------------------
>>
>> What could be the reason behind this? As far as I remember I have not
>> faced this problem in 3.3.3 version. Not that I regularly use parallel
>> version for energy minimization, but it would be nice to be able to
>> sort out this problem. The PR or MD run goes fine in the same
>> settings.
>>
>> The mdp file is:
>>
>> title = cg+steep energymin spce watbox
>> cpp = /lib/cpp
>> define = -DFLEXIBLE
>>
>> integrator = cg
>> nsteps = 5000
>> nstcgsteep = 1000
>> comm_mode = linear
>> nstcomm = 1
>>
>> emtol = 1.0
>> emstep = 0.01
>>
>> nstxout = 100
>> nstvout = 100
>> nstfout = 100
>> nstlog = 100
>> nstenergy = 100
>> energygrps = protein sol
>>
>> nstlist = 5
>> ns_type = grid
>> pbc = xyz
>> rlist = 0.9
>>
>> coulombtype = PME
>> rcoulomb = 0.9
>>
>> vdwtype = cut-off
>> rvdw = 1.2
>>
>> fourier_spacing = 0.12
>> pme_order = 6
>> ewald_rtol = 1e-6
>> optimize_fft = yes
>>
>> constraints = none
>>
>>
>>
>> Thanks,
>> Suman.