[CP2K:664] Re: MD simulation doesn't run very long

Teodoro Laino teodor... at gmail.com
Fri Feb 1 21:52:29 CET 2008


Hi Andrew,

no worries.. it's a quite well-known story.. the fact is that  
regarding compilers and libraries
cp2k is a little bit snobby ;-).. They must be able to handle  
*almost* perfectly the F90/95 (and possibly 2003) standard.
If you look in the old threads of this mailing list you will also  
find several other messages where people
had similar problems with PGI and intel.

Unfortunately not all compilers can compiler cp2k (and this mainly  
because it's 18 years that exists the F90 standard and
still there is not a robust compiler!!).. so check in the old  
messages, if you want to use intel or PGI, the versions of those
compilers properly compiling cp2k.

g95 generally does, though is not producing a very fast code.

Anyway you can be happy! Though you were not able to run cp2k with  
your input file, it is just CORRECT!
Maybe GMAX for ewald is a little bit too small ... and instead of  
using EWALD you can switch to the more
efficient SPME. But again all things are at the proper place! ;-)

The only thing I noticed is that you're still running an *old* cvs  
snapshot  (I hope that a real version will be soon not just a  
dream! ;-) )..
If you will do an update you will find that the keyword EACH is not  
there anymore..
Have a look at the proper message and of course feel free to ask if  
you need any clarification.

For honesty sake I must also say that you should not expect the same  
performance of other classical MD code when using CP2K
for classical runs (you can easily see up to a factor of 5-10 slower  
compared to the others available on the market).
We are aware of that, but that's the price you've to pay to be more  
general than just  doing biomolecular systems.
If you do just biomolecular what I usually suggest (and I do) is to  
equilibrate with other faster codes (to save time) and switching to
cp2k for doing QMMM (once the system is equilibrated with other codes  
using AMBER or CHARMM FF you will get the same
equilibration switching to cp2k..i.e. they are compatible).

Have a nice week-end,
Teo

On 1 Feb 2008, at 21:28, andrew... at eas.gatech.edu wrote:

>
> Ciao Teo,
>
> Well hmmm.  We had tried running the simulation on mpp2 at EMSL using
> another version of cp2k and that gave us the same error.  I think that
> version was compiled with pgi -- but we were also using the binary in
> someone else's directory so there could have been libraries that got
> their links messed up.
>
> I have read about the issues with gfortran.  I used it anyway because
> that was the only thing I could even get cp2k to compile with!  :)
> I'll try it again with g95.
>
> Thanks a lot for your help and running the file for me!!
>
> Cheers,
> Andrew
>
>
> On Feb 1, 2:11 pm, Teodoro Laino <teodor... at gmail.com> wrote:
>> Ciao Andrew,
>>
>>> lammps but cp2k ran into some connectivity issues (cp2k doesn't seem
>>> to find water molecules automatically across cell boundaries).  I
>>
>> this is not the case.. cp2k uses PBC also to identify the  
>> connectivity..
>> The point could be that water is a little bit tricky.. if the
>> geometry is a little bit distorted with hydrogen bonds is
>> anyway difficult to identify a proper water molecule as an
>> independent unity..
>> That's why is always better in my opinion to use an external topology
>> file as you did after..
>>
>>> fixed this, by making certain the water molecules were all in the
>>> image reference and defining the connectivity manually.  This didn't
>>> fix that specific problem though: the symptom is that the simulation
>>> runs for ~20-30 time steps and then the temperature increases
>>> dramatically and we get some atoms crash:
>>
>> I was running your input file for several time steps (see the
>> attached output file) and I don't get
>> any error..
>> I suppose gfortran is miscompiling cp2k (and I would not be
>> surprised.. as far as I know only very recent version of gfortran
>> can compile cp2k)..
>>
>> I would suggest you to use g95 (if you want to go for a free
>> compiler) or intel.
>>
>> I'm sorry but as you may have already read on this group cp2k is
>> quite demanding in terms of fortran compilers..
>> Hope this helps!
>> Teo
>>
>>  out.gz
>> 73KDownload
>>
>>
>>
>>
>>
>>> ******************************************************************** 
>>> **
>>> *********
>>>  ENSEMBLE TYPE
>>> =                                              nvt
>>>  STEP NUMBER
>>> =                                               26
>>>  TIME [FS]                    =
>>> 26.000000
>>>  CONSERVED QNTY               =
>>> -0.299834451059E+01
>>
>>>                                          INSTANTANEOUS         
>>> AVERAGES
>>>  CPU [S]                      =
>>> 0.22                 0.39
>>>  {E-E0}/{k_b*N_at}            =          0.396202120408E+02
>>> 0.335570376311E+01
>>>  POTENTIAL ENERGY[hartree]    =         -0.549743589253E+01
>>> -0.462863744377E+01
>>>  KINETIC ENERGY[hartree]      =          0.249463138359E+01
>>> 0.153631845774E+01
>>>  TEMPERATURE[K]               =
>>> 656.451              404.275
>>
>>> ******************************************************************** 
>>> **
>>> *********
>>
>>>  WARNING| Particles:     609    265 at distance [au]:     0.49549362
>>> less than:      0.95400000; increase EMAX_SPLINE.
>>
>>> The temperature was supposed to be 300 K, but as you can see it  
>>> blows
>>> up the last few time steps.
>>
>>> Here is a link to the files:
>>> http://shadow.eas.gatech.edu/~astack/water.tar.gz
>>
>>> Here is my input file, the commented out sections are different  
>>> things
>>> we tried to see if that made a difference.  The biggest thing that
>>> seemed to affect the number of time steps was the options affecting
>>> the spline.
>>
>>> &FORCE_EVAL
>>>   METHOD FIST
>>>   &MM
>>>     &FORCEFIELD
>>> !       &SPLINE
>>> !               EMAX_SPLINE 1000.0
>>> !               R0_NB [angstrom] 0.1
>>> !       &END SPLINE
>>>       &BEND
>>>          ATOMS H O H
>>>          KIND HARMONIC
>>>          K [kcalmol*rad^-2] 120
>>>          THETA0 [deg] 109.47
>>>        &END BEND
>>>        &BOND
>>>          ATOMS O H
>>>          KIND HARMONIC
>>>          K [kcalmol*angstrom^-2] 500
>>>          R0 [angstrom] 1.00
>>>        &END BOND
>>>       &NONBONDED
>>>         &LENNARD-JONES
>>>            ATOMS O O
>>>            EPSILON [kcalmol] 0.1848
>>> !           SIGMA [angstrom] 3.5532
>>>            SIGMA [angstrom] 3.1655
>>>            RCUT [angstrom] 10.0
>>>          &END LENNARD-JONES
>>>          &LENNARD-JONES
>>>            ATOMS O H
>>>            EPSILON [kcalmol] 0.0430
>>> !           SIGMA [angstrom] 1.7883
>>>            SIGMA [angstrom] 1.5932
>>>            RCUT [angstrom] 10.0
>>>          &END LENNARD-JONES
>>>          &LENNARD-JONES
>>>            ATOMS H H
>>>            EPSILON [kcalmol] 0.0100
>>> !           SIGMA [angstrom] 0.900
>>>            SIGMA [angstrom] 0.8018
>>>            RCUT [angstrom] 10.0
>>>          &END LENNARD-JONES
>>>       &END NONBONDED
>>>     &END FORCEFIELD
>>>     &POISSON
>>>       &EWALD
>>>         EWALD_TYPE ewald
>>>         ALPHA 0.5
>>>         GMAX 15
>>>       &END EWALD
>>>     &END POISSON
>>>   &END MM
>>>   &SUBSYS
>>>     &CELL
>>>       ABC 20 20 20
>>>       UNIT ANGSTROM
>>>       PERIODIC XYZ
>>>     &END CELL
>>>     &TOPOLOGY
>>>         CONN_FILE_NAME water.psf
>>>         CONNECTIVITY UPSF
>>>         COORD_FILE_NAME water.pdb
>>>         COORDINATE PDB
>>>     &END TOPOLOGY
>>>     &PRINT
>>>       &TOPOLOGY_INFO HIGH
>>>         FILENAME topology.inf
>>>         PDB_INFO
>>>         COMMON_ITERATION_LEVELS 1
>>>         UTIL_INFO
>>>         GENERATE_INFO
>>>       &END TOPOLOGY_INFO
>>>     &END PRINT
>>>   &END SUBSYS
>>> !  &RESCALE_FORCES
>>> !    MAX_FORCE 0.001
>>> !  &END
>>> !  &PRINT
>>> !       &FORCES
>>> !         FILENAME water.force
>>> !        &END FORCES
>>> !  &END PRINT
>>> &END FORCE_EVAL
>>> &GLOBAL
>>>   PROJECT water
>>>   RUN_TYPE MD
>>> !  PRINT_LEVEL HIGH
>>> &END GLOBAL
>>> &MOTION
>>>   &MD
>>>     ENSEMBLE NVT
>>>     STEPS 300
>>>     TIMESTEP 1
>>>     TEMPERATURE 300
>>> !    TEMP_TOL 1.0
>>>    &THERMOSTAT
>>>     &NOSE
>>>       LENGTH 1
>>>       YOSHIDA 3
>>>       MTS 2
>>>       TIMECON 500
>>>     &END NOSE
>>>    &END THERMOSTAT
>>>    &PRINT
>>>     &ENERGY HIGH
>>>      EACH 1
>>>     &END ENERGY
>>>    &END PRINT
>>>   &END MD
>>> &END MOTION
>>
>>>
>
> >




More information about the CP2K-user mailing list