[Wien] Error in parallel lapw1
Roberto Iglesias
roberto at uniovi.es
Thu Oct 16 13:06:26 CEST 2008
Thanks a lot for the fast reply, Peter. I've
verified what you say. Issuing now
x lapw0
x lapw1 -p -up
x lapw1 -p -dn
in a clean initialized test directory works just fine.
Using
runsp_lapw -p
from the terminal also works.
However, sending the job script with
runsp_lapw -p
gives the same lapw1 error as before. It seems
that the problem is in the script.
It's a PBS queuing system, and when submitting,
lapw0 works fine, but then something is wrong
with lapw1. I will try to work a little bit more
on the script and see if I find a work around.
Best regards!
Roberto
At 10:34 16/10/2008, you wrote:
>More than trivial !
>
>I'm sure, it does not work on your local computer either.
>The error says:
> > 'INILPW' - can't open unit: 18
> > 'INILPW' - filename: test.vsp
>but you searched and found:
> > The files test.vspup, test.vspdn ...
>
>When you have a spin-polarized case, you have to use
>
>x lapw1 -p -up (and -dn) (or for scf use runsp_lapw, not run_lapw).
>
>ROBERTO LUIS IGLESIAS PASTRANA schrieb:
> > Hello all!
> >
> > I'm trying to do a parallel calculation in a
> test system, which needs spin polarization. The
> machine is a 50 dual-node Xeon processor
> cluster. The Wien2k version is 08_3. I've
> searched the mailing list for problems similar
> to mine, but the few possible solutions I found do not help.
> > I use a small PBS script, based on the one
> appearing in the FAQ. If you need to take a
> look at it, let me know. The error is always:
> >
> > 'INILPW' - can't open unit: 18
> > 'INILPW' - filename: test.vsp
> > 'INILPW' - status: old form: formatted
> > 'LAPW1' - INILPW aborted unsuccessfully.
> >
> > The files test.vspup, test.vspdn and
> test.vsp_st are there and not empty. I don't see anything peculiar in them.
> >
> > There is a post in the mailing list that
> states that this problem got solved wuth a new
> .machines file, since the original one had only one node:
> >
> > http://zeus.theochem.tuwien.ac.at/pipermail/wien/2008-September/011427.html
> >
> > This is not my case, I think. My .machines file is:
> >
> > 1:yed16:1
> > 1:yed16:1
> > 1:yed15:1
> > 1:yed15:1
> > 1:yed14:1
> > 1:yed14:1
> > 1:yed13:1
> > 1:yed13:1
> > granularity:1
> > extrafine:1
> >
> > which from testpara_lapw would give:
> >
> > [iglesias at jaula02 test]$ testpara_lapw
> >
> > #####################################################
> > # TESTPARA #
> > #####################################################
> >
> > Test: LAPW1 in parallel mode (using .machines)
> > Granularity set to 1
> > Extrafine set
> >
> > klist: 500
> > machines: yed16 yed16 yed15 yed15 yed14 yed14 yed13 yed13
> > procs: 8
> > weigh(old): 1 1 1 1 1 1 1 1
> > sumw: 8
> > granularity: 1
> > weigh(new): 62 62 62 62 62 62 62 62
> >
> > Distribution of k-point (under ideal conditions)
> > will be:
> >
> > 1 : yed16(62) 62k
> > 2 : yed16(62) 62k
> > 3 : yed15(62) 62k
> > 4 : yed15(62) 62k
> > 5 : yed14(62) 62k
> > 6 : yed14(62) 62k
> > 7 : yed13(62) 62k
> > 8 : yed13(62) 62k
> > 9 : yed16(62) 1k
> > 10 : yed16(62) 1k
> > 11 : yed15(62) 1k
> > 12 : yed15(62) 1k
> >
> > It would therefore process 62 k-points in
> each of the 8 nodes (496 processes) and the
> remaining 4 one on each node whenever thay are free.
> > Without using the script, that is, issuing
> runsp_lapw -p from a terminal window, there are
> no problems. That should make use of the former
> .machines file. With the script the errors are
> as the one above for each lapw1 that is distributed to the nodes. In that case:
> >
> > [iglesias at jaula02 test]$ testpara1_lapw
> >
> > #####################################################
> > # TESTPARA1 #
> > #####################################################
> >
> > Wed Oct 15 18:53:22 CEST 2008
> >
> > lapw1para exited due to an ERROR
> > Check *.error files
> >
> > If I try from the terminal window:
> >
> > [iglesias at jaula02 test]$ lapw1 -p lapw1.def
> > LAPW1 - Error
> >
> > simply.
> >
> > On doing:
> >
> > x lapw1 -p
> >
> > it crashes again on each node, successively.
> >
> > Any suggestions? Thanks a lot in advance for your help!
> >
> > Cheers
> >
> > Roberto
> >
> >
> > _______________________________________________
> > Wien mailing list
> > Wien at zeus.theochem.tuwien.ac.at
> > http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>
>--
>
> P.Blaha
>--------------------------------------------------------------------------
>Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
>Phone: +43-1-58801-15671 FAX: +43-1-58801-15698
>Email: blaha at theochem.tuwien.ac.at WWW: http://info.tuwien.ac.at/theochem/
>--------------------------------------------------------------------------
>_______________________________________________
>Wien mailing list
>Wien at zeus.theochem.tuwien.ac.at
>http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
Roberto Iglesias Pastrana
Despacho 22
Área de Física Aplicada
Departamento de Física
Facultad de Ciencias
Universidad de Oviedo
C/ Calvo Sotelo, s/n 33007 Oviedo
Tel.: 34 (9) 85102898
34 (9) 85458136
Fax: 34 (9) 85103324
e-mail: roberto at uniovi.es
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20081016/b429b61d/attachment.html
More information about the Wien
mailing list