[Wien] Similar problem another user had, but I am having trouble understanding the solution

Peter Blaha pblaha at theochem.tuwien.ac.at
Tue Feb 19 17:10:59 CET 2013


Parallelization:

For your case I recommend    k-point parallelization, not 
mpi-parallelization.

The .machines file looks different for these two strategies.

k-parallel needs to be able to do passwordless login. It is well 
described in the UG how to set this up.

Am 18.02.2013 23:02, schrieb Joshua Davis:
> So I had already done calculations with Wien2k 11.1 in the directory
> before.  Because of certain reasons I had to update to 12.1.  I had
> compiled 12.1 with no error.  When I started doing this calculation I
> started with a spin polarized case with an effective U value of 4 eV
> with only 100 kpoints. When I started back up again I want to increase
> to 500 k points. So I tried to do the same but with 500 kpoints and i
> put on the parallel switch. That is were the trouble started.  at that
> point the only thing that showed up in my error files was  "error in
> LAPW1."   I had just compiled the program in parallel so I took off the
> parallel switch after I did clean_lapw, but I still have the same
> error.  So I took off the -orb switch. But, still the error persisted.
> So I tried to restart initialization program and do a regular
> calculation. (At this point it gave that error that happens if you try
> to do a regular calculation with spin polarized files.)  So maybe there
> were left over information even though I thought I cleaned it properly.
>
> The furthest the file  *.scf1up got up to (its the same for down):
>
>    K=   0.05556   0.05556   0.10000            1
> :RKM  : MATRIX SIZE 3245LOs: 140  RKM= 8.00  WEIGHT= 8.00  PGR:
>         EIGENVALUES ARE:
>         ********************************************************
>
> :KPT   :      NUMBER OF K-POINTS:    45
>
>
> and this shows up in *.output1up (just a portion):
>
>
>
> blebleble
>   LDA+U potential       added for atom type  2 L=  2 spin up
>
>    Orbital potential real part
>   M= -2        NaN        NaN        NaN        NaN        NaN
>   M= -1        NaN        NaN        NaN        NaN        NaN
>   M=  0        NaN        NaN        NaN        NaN        NaN
>   M=  1        NaN        NaN        NaN        NaN        NaN
>   M=  2        NaN        NaN        NaN        NaN        NaN
>
>    Orbital potential imaginary part
>   M= -2        NaN        NaN        NaN        NaN        NaN
>   M= -1        NaN        NaN        NaN        NaN        NaN
>   M=  0        NaN        NaN        NaN        NaN        NaN
>   M=  1        NaN        NaN        NaN        NaN        NaN
>   M=  2        NaN        NaN        NaN        NaN        NaN
>
>   LDA+U potential       added for atom type  3 L=  2 spin up
>
> However, I decide to move the *.struct, *.inorb, *.indm and the *.inst
> to another directory and I started over and I ran it without the
> parallel switch and it works now.
>
> I guess other information to add is that I did not compile as root or
> sudo because ifort is not sourced to root or sudo on my computer.
>
> Thanks Joshua Davis
>
>
> On Mon, Feb 18, 2013 at 11:50 AM, Peter Blaha
> <pblaha at theochem.tuwien.ac.at <mailto:pblaha at theochem.tuwien.ac.at>> wrote:
>
>     Your report is inconsistent.
>
>     After   lapw0   the steps   lapw1 -up and -dn follow.
>
>     In a spin-polarized case, this schould produce   *scf1up/dn and
>     *output1up/dn
>
>     Furthermore, the line
>
>      > :e__0001: OVERALL ENERGY PARAMETER IS    0.3393
>     proofs, that this is NOT the first iteration, but the in1(c) file
>     has been modified by lapw2.
>
>     There must be some message after lapw1 -up, and the error must
>     already be there  (cat *lapw1*.error)
>
>     Everything else are follow ups and do not really matter, until you
>     were able to produce a output1up/dn file which contains eigenvalues
>     for all your k-points.
>
>     Most likely you ahve problems finding the linearization energies....
>
>     Sphere radii ???
>
>     Am 18.02.2013 17:08, schrieb Joshua Davis:
>
>         Dear Wien2K users
>         I have been running a spin polarized calculation calculation
>         with 12.1
>         recently and I cam across this error that occured after LAPW0
>         first cycle:
>
>         forrtl: severe (174): SIGSEGV, segmentation fault occurred
>         Image              PC                Routine            Line
>             Source
>         lapw2              000000000043C83C  fermi_tetra_              519
>         fermi_tmp_.F
>         lapw2              000000000043BCDD  fermi_                    111
>         fermi_tmp_.F
>         lapw2              000000000047A9E7  MAIN__                    278
>         lapw2_tmp_.F
>         lapw2              0000000000403E9C  Unknown
>         Unknown  Unknown
>         libc.so.6          00002B49FD10C76D  Unknown
>         Unknown  Unknown
>         lapw2              0000000000403D99  Unknown
>         Unknown  Unknown
>         cp: cannot stat `.in.tmp': No such file or directory
>
>           >   stop error
>
>         This is similar to the thread here:
>         http://zeus.theochem.tuwien.__ac.at/pipermail/wien/2012-__October/017732.html
>         <http://zeus.theochem.tuwien.ac.at/pipermail/wien/2012-October/017732.html>
>
>         Where the answer may have been posted here:
>         http://zeus.theochem.tuwien.__ac.at/pipermail/wien/2012-__October/017741.html
>         <http://zeus.theochem.tuwien.ac.at/pipermail/wien/2012-October/017741.html>
>
>         However I am having trouble understanding how to utilize this
>         method.
>         Secondly I do not want to use  the fft packages that come with
>         Wien2k,
>         and rather use fftw3 as I want to run Wien2K in parallel.
>
>         I have a computer with Ubuntu 12.04 with two six-core Xeon
>         processors
>         and 48 gb of ram.  My Wien2K 12.1 was compiled with ifort 13 and
>         fftw 3.3.3.
>         my mpif90 was compiled with ifort 13 and MPICH 3.0.2
>         my fftw-3.3.3 library was compiled with ifort for any FORTRAN
>         portions.
>
>
>
>         other info:
>
>         This is all that is written in the *.scf1 file:
>
>
>                     ATOMIC SPHERE DEPENDENT PARAMETERS FOR ATOM  Sr1
>         :e__0001: OVERALL ENERGY PARAMETER IS    0.3393
>                     OVERALL BASIS SET ON ATOM IS LAPW
>
>
>         And this is all that is written in *.output1 file:
>
>
>         blebleble
>         0         P      TYPE LATTICE ASSUMED
>                     RELA-CALCULATION
>
>                     R-MT TIMES K-MAX IS 8.00
>                     MAX L IS 10     MAX L IN NONSPHERICAL MATRIXELEMENTS:  4
>            NUMBER OF ATOMS IS    9
>              0.09925   0.00000   0.00000
>             -0.00000   0.09925   0.00000
>             -0.00000  -0.00000   0.06423
>             10.07508   0.00000   0.00000
>              0.00000  10.07508   0.00000
>              0.00000   0.00000  15.56946
>
>         For both of them they just stop writing information.
>
>         Also, while this does seem similar to the other case previously
>         mentioned before, they do not have 'cp: cannot stat `.in.tmp':
>         No such
>         file or directory' at the end of _their_ error.
>
>
>         Any help would be appropriated, and I can print my *.struct file if
>         desired, but I think this post is getting long winded.
>
>         Josh Davis
>
>
>
>         _________________________________________________
>         Wien mailing list
>         Wien at zeus.theochem.tuwien.ac.__at
>         <mailto:Wien at zeus.theochem.tuwien.ac.at>
>         http://zeus.theochem.tuwien.__ac.at/mailman/listinfo/wien
>         <http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien>
>
>
>     --
>     Peter Blaha
>     Inst.Materials Chemistry
>     TU Vienna
>     Getreidemarkt 9
>     A-1060 Vienna
>     Austria
>     +43-1-5880115671 <tel:%2B43-1-5880115671>
>     _________________________________________________
>     Wien mailing list
>     Wien at zeus.theochem.tuwien.ac.__at
>     <mailto:Wien at zeus.theochem.tuwien.ac.at>
>     http://zeus.theochem.tuwien.__ac.at/mailman/listinfo/wien
>     <http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien>
>
>
>
>
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>

-- 
Peter Blaha
Inst.Materials Chemistry
TU Vienna
Getreidemarkt 9
A-1060 Vienna
Austria
+43-1-5880115671


More information about the Wien mailing list