[gmx-users] Error in BlueGene
Ricardo O. S. Soares
rsoares at fcfrp.usp.br
Mon Feb 4 17:00:56 CET 2013
Hello Cíntia,
On 02/01/2013 04:19 PM, Cintia C. Vequi-Suplicy wrote:
> Hello Ricardo,
>
> I tried to decrease the integration time but I got the same error.
> The system is running fine in my local core.
>
> I have two questions:
>
> 1) What do you mean by "I balance the domain decomposition for the
> larger number of core"?
When you run at your local hardware, the 8 cores handles your system
differently than the 1024, 2048 (or more) cores in bluegene. So you
must balance the number of cores with your system size. If you try to
run a small system in too many cores, an error may appear:
http://www.gromacs.org/Documentation/Errors#There_is_no_domain_decomposition_for_n_nodes_that_is_compatible_with_the_given_box_and_a_minimum_cell_size_of_x_nm
So you must decrease their number, or assign different numbers of cores
only for PME (mdrun options). There's also the scalability/performance
issue....
> 2) Are you running simulations at Bluegene P at Rice University?
>
Yes!
> Thank you,
> Cíntia
>
>
> On 30/01/2013 20:48, Ricardo Soares wrote:
>> Hello,
>>
>> before submitting to Bluegene, I always test the system in my local 8
>> core
>> cpu, and if it works, it will also work in Bluegene, as long as I
>> balance the
>> domain decomposition for the larger number of cores. If your system
>> insists in
>> exploding, even after energy minimization, you could try to decrease the
>> integration timestep (dt) to something lower than 2 fs, maybe to 1fs
>> or, in
>> some cases even 0.5 fs. Did you tried that?
>>
>> Cheers,
>>
>> Ricardo.
>>
>>
>>
>> On Wed, 30 Jan 2013 15:44:53 -0200, Cintia C. Vequi-Suplicy wrote
>>> Mark,
>>>
>>> Thank you for your answer.
>>>
>>> I did an energy minimization with the file em.mdp (below) and then I
>>> did a 40 ns simulation with the same md.mdp I send before. These
>>> two steps were done in my local cluster. After that I took the
>>> configuration for the simulation in the bluegene and I got these
>>> error. I notice that the step that the error occurs depends on the
>>> number of nodes I use. I am also including the end of the min.out file.
>>>
>>> I also tried others integrators for the energy minimization but I
>>> always get the same error.
>>>
>>> Thank you again,
>>> Cíntia
>>>
>>> -------------------------------------------------------
>>> em.mdp
>>>
>>> integrator = steep
>>> tinit = 0.0
>>> dt = 0.002
>>> nsteps = 5000000
>>> nstcomm = 1
>>> comm-grps = Other SOL
>>>
>>> nstxout = 100
>>> nstvout = 100
>>> nstfout = 0
>>> nstlog = 100
>>> nstenergy = 100
>>> nstxtcout = 100
>>> xtc_precision = 100
>>> xtc-grps =
>>> energygrps = Other SOL
>>>
>>> nstlist = 5
>>> ns_type = grid
>>> pbc = xyz
>>> rlist = 1.4
>>>
>>> coulombtype = reaction-field
>>> rcoulomb = 1.4
>>> epsilon_rf = 66
>>>
>>> vdw_type = cut-off
>>> rvdw = 1.4
>>> DispCorr = No
>>>
>>> constraints = none
>>> -------------------------------------------------------
>>> min.out
>>> ...
>>> Step= 5192, Dmax= 1.9e-11 nm, Epot= -1.79302e+06 Fmax= 6.65023e+02,
>>> atom= 22015
>>> Step= 5193, Dmax= 9.5e-12 nm, Epot= -1.79302e+06 Fmax= 6.65023e+02,
>>> atom= 22015
>>> Step= 5194, Dmax= 4.7e-12 nm, Epot= -1.79302e+06 Fmax= 6.65023e+02,
>>> atom= 22015
>>> Step= 5195, Dmax= 2.4e-12 nm, Epot= -1.79302e+06 Fmax= 6.65023e+02,
>>> atom= 22015
>>> Step= 5196, Dmax= 2.8e-12 nm, Epot= -1.79302e+06 Fmax= 6.65023e+02,
>>> atom= 22015
>>> Step= 5197, Dmax= 1.4e-12 nm, Epot= -1.79302e+06 Fmax= 6.65023e+02,
>>> atom= 22015
>>> Step= 5198, Dmax= 1.7e-12 nm, Epot= -1.79302e+06 Fmax= 6.65023e+02,
>>> atom= 22015
>>> Stepsize too small, or no change in energy.
>>> Converged to machine precision,
>>> but not to the requested precision Fmax < 10
>>> You might need to increase your constraint accuracy, or turn
>>> off constraints alltogether (set constraints = none in mdp file)
>>>
>>> writing lowest energy coordinates.
>>>
>>> Steepest Descents converged to machine precision in 5199 steps,
>>> but did not reach the requested Fmax < 10.
>>> Potential Energy = -1.79302464181879e+06
>>> Maximum force = 6.65022934447791e+02 on atom 22015
>>> Norm of force = 4.90065871184878e+00
>>>
>>> gcq#0: Thanx for Using GROMACS - Have a Nice Day
>>> -------------------------------------------------------
>>>
>>>>
>>>> -------------------------------------------------------
>>>> md.mdp file
>>>>
>>>>
>>>> integrator = md
>>>> tinit = 0.0
>>>> dt = 0.002
>>>> nsteps = 5000000
>>>> nstcomm = 5
>>>> comm-grps = Other SOL
>>>>
>>>> nstxout = 1000
>>>> nstvout = 1000
>>>> nstfout = 0
>>>> nstlog = 1000
>>>> nstenergy = 1000
>>>> nstxtcout = 1000
>>>> xtc_precision = 1000
>>>> xtc-grps =
>>>> energygrps = Other SOL
>>>>
>>>> nstlist = 5
>>>> ns_type = grid
>>>> pbc = xyz
>>>> rlist = 1.4
>>>>
>>>> coulombtype = reaction-field
>>>> rcoulomb = 1.4
>>>> epsilon_rf = 66
>>>>
>>>> vdw_type = cut-off
>>>> rvdw = 1.4
>>>> DispCorr = No
>>>>
>>>> tcoupl = v-rescale
>>>> tc-grps = Other SOL
>>>> tau_t = 0.4 0.4
>>>> ref_t = 296 296
>>>>
>>>> Pcoupl = berendsen
>>>> Pcoupltype = semiisotropic
>>>> tau_p = 0.2 0.2
>>>> compressibility = 4.51e-5 4.51e-5
>>>> ref_p = 1.0 1.0
>>>>
>>>> constraints = hbonds
>>>> constraint_algorithm = lincs
>>>>
>>> --
>>> 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
>
More information about the gromacs.org_gmx-users
mailing list