[Wien] Problems with parallel hybrid functional calculation

Peter Blaha pblaha at theochem.tuwien.ac.at
Wed Mar 4 07:39:48 CET 2015


I do not understand your    .machines file.

It is the worst possible setting you can make:

a) with a 2x2x2 k-mesh you have at most 8 k-points (or in your case 4 ??, depending on symmetry),
so using more than 4 lines is absolute MEANINGLESS.

b) The mpi-code has to use a different diagonalization routine, which is a priori
almost 2 times slower than the sequential code. This when using 2 cores mpi-parallelization
you are probably even slower than in sequential mode.

c) As Fabien Tran pointed out, you should definitely test it with 1 kpoint first !!!!,
    or at least use the "reduced k-mesh" (1 k).

d) As far as I can see, the system is metallic. Hybrid-DFT is quite bad for "real" metals.

e) with 24 cores and 4 k-points a reasonable   .machines file should have 4 lines with

1:localhost:6


Are you using Wien2k_14.2 ???

What is your $SCRATCH ???   Use SCRATCH=./   for security.

Am 04.03.2015 um 01:21 schrieb Paul Fons:
>
> Hi All,
>    I am trying to run a hybrid functional calculation on a 96 atom system using PBE0.  I successfully used a PBE functional using the test machines file below that is a
> hybrid k-point/MPI parallel job.  While I posted a day ago about a failure due to missing ibz fbz files when I attempted to use the web interface, this time, I strictly
> used the terminal.  The output is shown below.  There were no problems with the fbz/ibz file generation, however, the run still terminated with an error.  The error was
> that case.vectorhf_old was not found.  I am not sure where this file show have been generated or why it was not.  I expect this calculation to take some time to run and am
> planning on using 96-cores for the real PBE0 calculation, but at the moment, I am in troubleshooting mode.  Does anyone have any idea of what might have gone wrong.  I used
> a shifted 2x2x2 MP grid and 770 bands as I have 1536 electrons.
>
> Thanks,
>       Paul Fons
>
>
> (.machines file)
>
> lapw0:localhost:24
> 1:localhost:2
> 1:localhost:2
> 1:localhost:2
> 1:localhost:2
> 1:localhost:2
> 1:localhost:2
> 1:localhost:2
> 1:localhost:2
> 1:localhost:2
> 1:localhost:2
> 1:localhost:2
> 1:localhost:2
> #1:draco-ib:12
> granularity:1
> extrafine:1
>
>
> TERMINAL OUTPUT
>
> matstud at gemini.a04.aist.go.jp <mailto:matstud at gemini.a04.aist.go.jp>:/usr/local/share/wien2k/Fons/aCGT>init_hf_lapw
> :GAP  : -99999.   Ry = -9999.    eV  ( metallic )
>           Bandranges (emin - emax) and occupancy:
> :BAN00759: 759    0.353028    0.368374  2.00000000
> :BAN00760: 760    0.354099    0.369025  2.00000000
> :BAN00761: 761    0.357944    0.370696  2.00000000
> :BAN00762: 762    0.360732    0.374369  2.00000000
> :BAN00763: 763    0.362890    0.377049  2.00000000
> :BAN00764: 764    0.366857    0.378864  2.00000000
> :BAN00765: 765    0.370671    0.382529  2.00000000
> :BAN00766: 766    0.374310    0.386296  2.00000000
> :BAN00767: 767    0.379037    0.388138  2.00000000
> :BAN00768: 768    0.383069    0.392822  1.61356106
> :BAN00769: 769    0.387515    0.396096  0.38643894
> :BAN00770: 770    0.392627    0.399987  0.00000000
> :BAN00771: 771    0.396049    0.406404  0.00000000
> :BAN00772: 772    0.399687    0.414211  0.00000000
> :BAN00773: 773    0.410564    0.422211  0.00000000
> :BAN00774: 774    0.419101    0.430560  0.00000000
>          Energy to separate low and high energystates:   -0.09012
>
>
> :NOE  : NUMBER OF ELECTRONS          =1536.000
>
> :FER  : F E R M I - ENERGY(TETRAH.M.)=   0.3906577224
> You must set NBAND to at least  NB_occ + 1  and you have 1536.00 electrons
> edit aCGT.inhf …   (SET NUMBER OF BANDS TO 770)
> PS: For very accurate calc. and large NBAND you may have to increase EMAX in aCGT.in1 by hand
>
> Prepare k-mesh for HF. Use identical mesh and shift for IBZ and FBZ
> Do you want to use a REDUCED k-mesh for HF (saving cpu-time) (Y/n) ?
> n
>   This script runs   x kgen   twice and generates equivalent meshes in the
>   IBZ and FBZ.
>
>
>
> KGEN ENDS
> KGEN ENDS
> How many k-points in full BZ?
> If you type 0 you can give 3 numbers for nx,ny,nz
> 0
> How many in x?
> 2
> How many in y?
> 2
> How many in z?
> 2
> Do you want to shift? (yes=1, no=0)
> 1
>             1  symmetry operations without inversion
>   inversion added (non-spinpolarized non-so calculation)
>    NUMBER OF K-POINTS IN WHOLE CELL: (0 allows to specify 3 divisions of G)
>   length of reciprocal lattice vectors:   0.246   0.245   0.247   0.000   0.000   0.000
>    Specify 3 mesh-divisions (n1,n2,n3):
>    Shift of k-mesh allowed. Do you want to shift: (0=no, 1=shift)
>             4  k-points generated, ndiv=           2           2           2
> KGEN ENDS
> 0.002u 0.008s 0:00.07 0.0%0+0k 24+128io 0pf+0w
>             1  symmetry operations without inversion
>    NUMBER OF K-POINTS IN WHOLE CELL: (0 allows to specify 3 divisions of G)
>   length of reciprocal lattice vectors:   0.246   0.245   0.247   0.000   0.000   0.000
>    Specify 3 mesh-divisions (n1,n2,n3):
>    Shift of k-mesh allowed. Do you want to shift: (0=no, 1=shift)
>             8  k-points generated, ndiv=           2           2           2
> KGEN ENDS
> 0.000u 0.008s 0:00.39 0.0%0+0k 72+40io 0pf+0w
> aCGT.in0_grr and aCGT.inhf and hf-kmesh prepared
> Now do the hybrid calculation:   run_lapw -hf  ...
> matstud at gemini.a04.aist.go.jp <mailto:matstud at gemini.a04.aist.go.jp>:/usr/local/share/wien2k/Fons/aCGT>run_lapw -hf -p
>   LAPW0 END
>   LAPW0 END
>   LAPW1 END
> mv: cannot stat `aCGT.vector': No such file or directory
>   LAPW1 END
>   LAPW1 END
>   LAPW1 END
>   LAPW1 END
> mv: cannot stat `aCGT.vectorhf_old': No such file or directory
> forrtl: severe (24): end-of-file during read, unit 1001, file /usr/local/share/wien2k/Fons/aCGT/031
> Image              PC                Routine            Line        Source
> lapw2c             0000000000648027  Unknown               Unknown  Unknown
> lapw2c             0000000000669FF3  Unknown               Unknown  Unknown
> lapw2c             0000000000482FEE  outp_                     180  outp.f
> lapw2c             00000000004701EC  l2main_                  2125  l2main_tmp_.F
> lapw2c             000000000047CEA5  MAIN__                    607  lapw2_tmp_.F
> lapw2c             000000000040407E  Unknown               Unknown  Unknown
> libc.so.6          000000343F61ED5D  Unknown               Unknown  Unknown
> lapw2c             0000000000403F89  Unknown               Unknown  Unknown
>
>  >   stop error
>
>
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> SEARCH the MAILING-LIST at:  http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
>

-- 
-----------------------------------------
Peter Blaha
Inst. Materials Chemistry, TU Vienna
Getreidemarkt 9, A-1060 Vienna, Austria
Tel: +43-1-5880115671
Fax: +43-1-5880115698
email: pblaha at theochem.tuwien.ac.at
-----------------------------------------


More information about the Wien mailing list