[Pw_forum] overlap matrix

Paolo Giannozzi p.giannozzi at gmail.com
Mon Jan 22 10:40:20 CET 2018


On Sun, Jan 21, 2018 at 6:12 AM, Andrew Supka <supka1ar at cmich.edu> wrote:


> If you haven't compiled with scalapack northo will always be set to 1.
>

I am not sure about this: in QE there is an alternative algorithm for
parallel dense-matrix diagonalization. Anyway: overlaps are not printed
when parallel dense-matrix diagonalization is used (it's just not
implemented, and presumably it never was in this case: see routine
"pprojwave").  Implementing it requires to collect the various pieces of
matrices that are distributed across processors. Setting "-northo=1" should
do the job (in addition to slowing down it a bit).

>
> Andrew Supka
> Central Michigan University
>
> On Jan 20, 2018 8:33 PM, "Negar Ashari Astani" <negarastani at gmail.com>
> wrote:
>
> Dear Kevin,
>
> Could be! I figured out that whenever I run with only one processor
> everything is fine. However, the parallel implementation does not write the
> overlap matrix. Also the output format is different. So I ended up running
> projwfc.x only on one processor! Maybe there is something wrong with the
> compilation and MKL libraries as you said!
>
> Many thanks,
>
> Negar
>
>
> --
> Negar Ashari-Astani, PhD
> Sharif university of technology,
> Department of Physics
> Room 624
> Tel: (+98)2166164504 <+98%2021%206616%204504>
>
> _______________________________________________
> 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, fax +39-0432-558222
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20180122/412cd27c/attachment.html>


More information about the users mailing list