[Pw_forum] GIPAW memory problem?

Aleksander Jaworski aleksander.jaworski at mmk.su.se
Wed May 21 18:27:29 CEST 2014


Dear Davide,

Using SVN versions of QE and GIPAW my glass nmr inputs are passing
initialization step finally, but then I'm getting an error:

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
     Error in routine cdiaghg (118):
     S matrix not positive definite

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

I have checked with '-ndiag 1' flag and it is same.

Do you have any ideas how to proceed further?
Seems that I'm really unlucky with my system or/and machine:):)

greetings,
Aleksander





full nmr output below:


 Program QE v.5.1rc2 starts on 20May2014 at  7: 3:26 

     This program is part of the open-source Quantum ESPRESSO suite
     for quantum simulation of materials; please cite
         "P. Giannozzi et al., J. Phys.:Condens. Matter 21 395502 (2009);
          URL http://www.quantum-espresso.org", 
     in publications or presentations arising from this work. More details
at
     http://www.quantum-espresso.org/quote

     Parallel version (MPI), running on     1 processors

     ***** This is GIPAW svn revision unknown      *****

     Parallelizing q-star over 1 images
     Message from routine gipaw_readin:
     *** isolve is obsolete, use diagonalization instead ***
     Message from routine gipaw_readin:
     *** iverbosity is obsolete, use verbosity instead ***

   Info: using nr1, nr2, nr3 values from input

   Info: using nr1s, nr2s, nr3s values from input

     IMPORTANT: XC functional enforced from input :
     Exchange-correlation      =  SLA  PW   PBX  PBC ( 1  4  3  4 0)
     Any further DFT definition will be discarded
     Please, verify this is what you really want

               file Sc.pbe-spn-kjpaw_psl.0.2.3.UPF: wavefunction(s)  3P 3D
renormalized
     WARNING: atomic wfc #  3 for atom type 2 has zero norm
     WARNING: atomic wfc #  4 for atom type 2 has zero norm
     WARNING: atomic wfc #  3 for atom type 3 has zero norm
     WARNING: atomic wfc #  4 for atom type 3 has zero norm

     G-vector sticks info
     --------------------
     sticks:   dense  smooth     PW     G-vecs:    dense   smooth      PW
     Sum        8109    3249    885               550217   138985   20341


     Subspace diagonalization in iterative solution of the eigenvalue
problem:
     a serial algorithm will be used

     GIPAW projectors -----------------------------------------------
     atom=  Sc   l=0  rc=    1.3000  rs=    0.8667
     atom=  Sc   l=0  rc=    1.3000  rs=    0.8667
     atom=  Sc   l=1  rc=    1.5000  rs=    1.0000
     atom=  Sc   l=1  rc=    1.8000  rs=    1.2000
     atom=  Sc   l=2  rc=    1.6000  rs=    1.0667
     atom=  Sc   l=2  rc=    1.6000  rs=    1.0667
     projs nearly linearly dependent: l=2  n1,n2= 1, 2  s=  0.99858295

     atom=  Al   l=0  rc=    2.0000  rs=    1.3333
     atom=  Al   l=1  rc=    2.0000  rs=    1.3333
     atom=  Al   l=0  rc=    2.0000  rs=    1.3333
     atom=  Al   l=1  rc=    2.0000  rs=    1.3333
     atom=  Al   l=2  rc=    2.0000  rs=    1.3333
     projs nearly linearly dependent: l=0  n1,n2= 1, 2  s=  0.99761080
     projs nearly linearly dependent: l=1  n1,n2= 1, 2  s=  0.99948100

     atom=  Si   l=0  rc=    2.0000  rs=    1.3333
     atom=  Si   l=1  rc=    2.0000  rs=    1.3333
     atom=  Si   l=0  rc=    2.0000  rs=    1.3333
     atom=  Si   l=1  rc=    2.0000  rs=    1.3333
     atom=  Si   l=2  rc=    2.0000  rs=    1.3333
     projs nearly linearly dependent: l=0  n1,n2= 1, 2  s=  0.99373142
     projs nearly linearly dependent: l=1  n1,n2= 1, 2  s=  0.99879796

     atom=  O    l=0  rc=    1.3000  rs=    0.8667
     atom=  O    l=0  rc=    1.3000  rs=    0.8667
     atom=  O    l=1  rc=    1.3000  rs=    0.8667
     atom=  O    l=1  rc=    1.3000  rs=    0.8667
     projs nearly linearly dependent: l=1  n1,n2= 1, 2  s=  0.99505596
     -----------------------------------------------------------------

     alpha_pv=     96.5280 eV

     GIPAW job: nmr                                                       
                    
     NMR macroscopic correction: yes
             0.6667          0.0000          0.0000
             0.0000          0.6667          0.0000
             0.0000          0.0000          0.6667


     Largest allocated arrays     est. size (Mb)     dimensions
        KS wavefunctions at k          26.58 Mb     (   17422,  100)
        KS wavefunctions at k+q        26.58 Mb     (   17422,  100)
        First-order wavefunctions     265.84 Mb     (   17422,  100, 10)
        First-order wavefunct (US)    159.50 Mb     (   17422,  100,  6)
        Charge/spin density             2.85 Mb     (  373248,    1)
        Induced current                25.63 Mb     (  373248,    3,3,1)
        Induced magnetic field         25.63 Mb     (  373248,    3,3,1)
        NL pseudopotentials            71.78 Mb     (   17422,  270)
        GIPAW NL terms                 95.70 Mb     (   17422,  360)

     Computing the magnetic susceptibility     isolve=1    ethr= 
0.1000E-13
     k-point #    1 of    36      pool #  1    cpu time:      39.1


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
     Error in routine cdiaghg (118):
     S matrix not positive definite

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

     stopping ... 






On Wed, 14 May 2014 13:55:57 +0200, Davide Ceresoli
<davide.ceresoli at istm.cnr.it> wrote:
> Dear Aleksander,
>      I've updated 'configure'. Do 'svn update' in qe-gipaw directory.
> I've tested with intel xe2011, intel xe2013 and gfortran-4.8 and
> both openmpi-1.6.x and mpich-3.x.
> 
> Let us know!
> 
> Best,
>      Davide
> 
> 
> 
> On 05/13/2014 08:41 PM, Aleksander Jaworski wrote:
>> Thank you very much Davide for your reply and suggestions.
>>
>> I'm using gfortran and mpif90 compilers.
>>
>> I have tried to run gipaw.x with '-ndiag 1' but is not changing
anything.
>>
>> I have compiled SVN version of QE and then using
>>
>> 'svn checkout svn://cvs.qe-forge.org/scmrepos/svn/qe-gipaw/trunk
>> qe-gipaw'
>>
>> downloaded SVN version of gipaw. But when I'm trying to compile it
error
>> occurs:
>>
>>
>> checking quantum-Espresso version... 5.1rc2
>> configure: error: Cannot compile against this version of
quantum-Espresso
>>
>> how could I go around that?
>>
>> To be honest, I'm not fully aware of the I/O bound, and I don't know
how
>> to identify it as a bottleneck.
>>
>> Thanks for checking my inputs. Fact that they are running on your
machine
>> means something fishy with my installation.
>>
>> best,
>> Aleksander
>>
>>
>>
>> On Tue, 13 May 2014 14:01:43 +0200, Davide Ceresoli
>> <davide.ceresoli at istm.cnr.it> wrote:
>>> Dear Aleksander,
>>>       the gipaw behavior you reported is clearly odd. The
initialization
>>> phase should last few seconds and do not consume much memory.
>>>
>>> I've just tested your input files and they work on my machine.
>>> In your case, GIPAW appears to be stuck while reading the charge
>>> density or in check_para_diag routine. Could you have an I/O problem?
>>> Which compiler/MPI are you using?
>>>
>>> Could you try gipaw with '-ndiag 1'? or, could you try the SVN
>>> version (both QE and GIPAW)?
>>>
>>> Let me know if nothing can work.
>>>
>>> Best,
>>>       Davide
>>>
>>>
>>>
>>>
>>> On 05/12/2014 05:18 PM, Aleksander Jaworski wrote:
>>>> Dear QE users and developers,
>>>>
>>>>
>>>> I'm experiencing issues with running gipaw.x on the glass structures
>>>> which
>>>> have been created from the classical MD simulations trajectories.
>>>> On the glass pw.x is running smoothly and converging properly.
Gipaw.x
>> is
>>>> starting, but showing very strange behavior in terms of memory
handling
>>>> and
>>>> never terminating the job.
>>> _______________________________________________
>>> Pw_forum mailing list
>>> Pw_forum at pwscf.org
>>> http://pwscf.org/mailman/listinfo/pw_forum
>>



More information about the users mailing list