[Pw_forum] Problems computing cholensky

Paolo Giannozzi p.giannozzi at gmail.com
Tue Feb 27 11:28:27 CET 2018


On Mon, Feb 26, 2018 at 10:26 AM, Laurens Siemons <laurenssiemons at hotmail.be
> wrote:


>  Does anybody have a suggestion why it does work on 1 node with 20 cores,
> but fails when I try to increase my nodes?
>

The final results do not depend upon the number of processors (modulo minor
numerical differences, within the convergence threshold) but intermediate
results may depend, due to a different starting point and to small
numerical differences. This may unfortunately lead some calculations that
are "on the brink" to fail. I have no evidence that this is related to a
bug or to any other easily solvable problem.

The "Cholesky" problem signals that the overlap matrix is not positive
definite (has a zero or negative eigenvalue). When it happens, it is
invariably due either to a badly wrong structure (not the case here) or to
USPP/PAW with small negative values of augmentation charges. It is a known
problem and there is no simple solution available.

Paolo

Thanks in advance,
>
> Laurens Siemons
>
>
> ------------------------------
> *Van:* pw_forum-bounces at pwscf.org <pw_forum-bounces at pwscf.org> namens
> Paolo Giannozzi <p.giannozzi at gmail.com>
> *Verzonden:* zondag 25 februari 2018 8:33
> *Aan:* PWSCF Forum
> *Onderwerp:* Re: [Pw_forum] Problems computing cholensky
>
> You can use cell_parameters together with A or celldm(1).
>
> On Sat, Feb 24, 2018 at 8:53 PM, Manu Hegde <mhegde at uwaterloo.ca> wrote:
>
> Hi,
> I do not know much about your system but looking quickly into the crystal
> structure there is something that might be causing problem. Looks like you
> have set ibrav=0, in that case you have to use card cell_parameters. A=xx
> not required. Also double check ypur system with xcrysden before starting
> the calculations.
> Manu
> (SFU)
>
> On Fri, Feb 23, 2018 at 10:19 AM, <elchatz at auth.gr> wrote:
>
> Hello Laurens Siemons,
>
> Although I am not one of the experts, I had the same problem in one of
> the scf runs I was doing for a GW calculation. Because of the high
> number of bands and ecutwfc that I needed to use and in order to get
> any results, I had to run the simulation on 100 cores. The strange
> thing for me also was that the first one I tried run, but then nothing
> again. After a few weeks of trying I was notified by our cluster
> services that I should not use more than 60 cores as the I/O
> operations that are done by QE were too high and the disk could not
> cope. I gave up GW since then, but if there is a solution to this
> problem, I would like to hear it too :S
>
>
> Eleni
>
>
> Quoting Laurens Siemons <laurenssiemons at hotmail.be>:
>
> > Dear all,
> >
> >
> > I'm a master student chemistry and I'm using QE (v. 6.1) for a relax
> > calculation of a rutile 101 slab with a vacuum above it.
> >
> > I'm getting the famous error:
> >
> >
> >
> > %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
> %%%%%%%%%%%%%%%%%%
> >      Error in routine  cdiaghg (161):
> >       problems computing cholesky
> >
> > %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
> %%%%%%%%%%%%%%%%%%
> >
> > I've read almost every related topic on the forum that I could find
> > and I tried a lot already to overcome this, like:
> > - changing values for ecutwfc and ecutrho
> > - changing mixing_beta
> > - changing functionals
> > - Tried to run the calcualtion with other input files (anatase 101,
> 001...)
> > - Changed diagonalization to 'cg' (In this case it calculates some
> > itterations but then crashes with the error: 'Error in routine
> > c_bands (1): >> too many bands are not converged')
> >
> > Nothing seems to help and I'm out of options... I even tried to run
> > a calculation of my predecessor (that has succeeded in the past) but
> > this also failed (he used an older version of QE though...).
> >
> > I'm postig my input file at the end here and I really hope somebody
> > can help me.
> >
> > Kind Regards,
> > Laurens Siemons
> >
> > &CONTROL
> >   calculation = 'relax'
> >   restart_mode = "from_scratch",
> >   prefix       = "testen",
> >   pseudo_dir =
> > '/data/antwerpen/204/vsc20442/pseudo/pslibrary.1.0.0/wc/PSEU
> DOPOTENTIALS'
> >   outdir = '/data/antwerpen/204/vsc20442'
> >   nstep = 100
> > /
> > &SYSTEM
> >   ibrav = 0
> >   A =    4.59631
> >   nat = 36
> >   ntyp = 2
> >   ecutwfc = 60
> >   ecutrho = 600
> > /
> > &ELECTRONS
> >   electron_maxstep = 300
> >   mixing_beta = 0.10
> >   conv_thr =  1.0d-8
> >   mixing_mode = 'local-TF'
> >   diago_thr_init = 1e-4
> > /
> > &IONS
> >   ion_dynamics = 'bfgs'
> >   ion_positions = 'default'
> > /
> > CELL_PARAMETERS {alat}
> >   1.000000000000000   0.000000000000000   0.640859733133753
> >   0.000000000000000   2.000000000000000   0.000000000000000
> >   0.000000000000000   0.000000000000000   3.845158398802518
> > ATOMIC_SPECIES
> >    O   15.99900   O.wc-n-kjpaw_psl.1.0.0.UPF
> >   Ti   47.86700   Ti.wc-spn-kjpaw_psl.1.0.0.UPF
> > ATOMIC_POSITIONS {crystal}
> > Ti  -0.000000000000000  -0.000000000000000   0.075000000000000 0 0 0
> > Ti  -0.000000000000000  -0.000000000000000   0.408333333333333
> > Ti  -0.000000000000000   0.500000000000000   0.241666666666667
> > Ti  -0.000000000000000  -0.000000000000000   0.241666666666667
> > Ti  -0.000000000000000   0.500000000000000   0.075000000000000 0 0 0
> > Ti  -0.000000000000000   0.500000000000000   0.408333333333333
> > Ti   0.500000000000000   0.250000000000000   0.075000000000000 0 0 0
> > Ti   0.500000000000000   0.250000000000000   0.408333333333333
> > Ti   0.500000000000000   0.750000000000000   0.241666666666667
> > Ti   0.500000000000000   0.250000000000000   0.241666666666667
> > Ti   0.500000000000000   0.750000000000000   0.075000000000000 0 0 0
> > Ti   0.500000000000000   0.750000000000000   0.408333333333333
> >  O   0.304303000000000   0.152151500000000   0.024282833333333 0 0 0
> >  O   0.304303000000000   0.152151500000000   0.357616166666667
> >  O   0.304303000000000   0.652151500000000   0.190949500000000
> >  O   0.304303000000000   0.152151500000000   0.190949500000000
> >  O   0.304303000000000   0.652151500000000   0.024282833333333 0 0 0
> >  O   0.304303000000000   0.652151500000000   0.357616166666667
> >  O   0.695697000000000   0.347848500000000   0.459050500000000
> >  O   0.695697000000000   0.347848500000000   0.292383833333333
> >  O   0.695697000000000   0.847848500000000   0.125717166666667 0 0 0
> >  O   0.695697000000000   0.347848500000000   0.125717166666667 0 0 0
> >  O   0.695697000000000   0.847848500000000   0.459050500000000
> >  O   0.695697000000000   0.847848500000000   0.292383833333333
> >  O   0.804303000000000   0.097848500000000   0.024282833333333 0 0 0
> >  O   0.804303000000000   0.097848500000000   0.357616166666667
> >  O   0.804303000000000   0.597848500000000   0.190949500000000
> >  O   0.804303000000000   0.097848500000000   0.190949500000000
> >  O   0.804303000000000   0.597848500000000   0.024282833333333 0 0 0
> >  O   0.804303000000000   0.597848500000000   0.357616166666667
> >  O   0.195697000000000   0.402151500000000   0.125717166666667 0 0 0
> >  O   0.195697000000000   0.402151500000000   0.459050500000000
> >  O   0.195697000000000   0.902151500000000   0.292383833333333
> >  O   0.195697000000000   0.402151500000000   0.292383833333333
> >  O   0.195697000000000   0.902151500000000   0.125717166666667 0 0 0
> >  O   0.195697000000000   0.902151500000000   0.459050500000000
> > K_POINTS {automatic}
> > 4 4 6 1 1 1
>
>
>
> --
> Dr. Eleni Chatzikyriakou
> Computational Physics lab
> Aristotle University of Thessaloniki
> elchatz at auth.gr - tel:+30 2310 998109
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://pwscf.org/mailman/listinfo/pw_forum
>
>
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://pwscf.org/mailman/listinfo/pw_forum
>
>
>
>
> --
> Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche,
> Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
> Phone +39-0432-558216 <+39%200432%20558216>, fax +39-0432-558222
> <+39%200432%20558222>
>
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://pwscf.org/mailman/listinfo/pw_forum
>


-- 
Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche,
Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
Phone +39-0432-558216 <+39%200432%20558216>, fax +39-0432-558222
<+39%200432%20558222>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20180227/55a4ce30/attachment.html>


More information about the users mailing list