[Pw_forum] "nproc_ortho > 1 not yet implemented" in projwfc.x run

H. Lee hjunlee at gmail.com
Fri Mar 13 18:19:42 CET 2015


Dear Prof. Paolo:

Thank you very much for your kindly reply.

> does this happen also if you do not specify anything for
> nproc_ortho (or nproc_diag, they are the same), or if you
> specify nproc_ortho=1, in the command line for projwfc.x?

I didn't specify nproc_ortho. It seems that mp_startup or something
automatically set this value according to the number of cores used in
calculation.
In my case, 6 cores set nproc_ortho=1, but 8 cores set nproc_ortho not
equal to 1.

>It is possible to run most calculations done by projwfc.x
>using distributed matrices and parallel diagonalization. This
>is useful for large systems where Nband*Nband matrices may be
>very large and not fit into memory.

Yes, I know. To my knowledge, currently only projwave without noncollinear
case has been implemented to use distributed matrices.

>This possibility however does not exist for case "tdosinboxes".
>Likely there is no need to stop the code, though. Just modify
>the part of code listed below and try:
>  IF ( tdosinboxes ) THEN
>     IF( nproc_ortho > 1 ) THEN
>        CALL errore ('do_projwfc', 'nproc_ortho > 1 not yet
>implemented', 1)
>     ELSE
>        CALL projwave_boxes (filpdos, filproj, n_proj_boxes, irmin,
>irmax, plotboxes)
>     ENDIF
>  ELSE
>...
>  ENDIF

Yes. I also worked around this problem by modifying this part.

Regards.

On Fri, Mar 13, 2015 at 3:06 PM, Paolo Giannozzi <paolo.giannozzi at uniud.it>
wrote:

> On Thu, 2015-03-12 at 16:26 +0100, H. Lee wrote:
>
> > When I set tdosinboxes = .true. and use multi-cores,
> > projwfc.x run in QE v5.1.1 compiled with Scalapack
> > results in the following error:
> > nproc_ortho > 1 not yet implemented
>
> does this happen also if you do not specify anything for
> nproc_ortho (or nproc_diag, they are the same), or if you
> specify nproc_ortho=1, in the command line for projwfc.x?
>
> It is possible to run most calculations done by projwfc.x
> using distributed matrices and parallel diagonalization. This
> is useful for large systems where Nband*Nband matrices may be
> very large and not fit into memory.
>
> This possibility however does not exist for case "tdosinboxes".
> Likely there is no need to stop the code, though. Just modify
> the part of code listed below and try:
>   IF ( tdosinboxes ) THEN
>      IF( nproc_ortho > 1 ) THEN
>         CALL errore ('do_projwfc', 'nproc_ortho > 1 not yet
> implemented', 1)
>      ELSE
>         CALL projwave_boxes (filpdos, filproj, n_proj_boxes, irmin,
> irmax, plotboxes)
>      ENDIF
>   ELSE
> ...
>   ENDIF
>
> Paolo
>
> >
> > But, to my understanding, nproc_ortho is related to the
> > parallelization in iterative diagonalization through distributed
> > matrices using Scalapack.
> >
> >
> > Thus, the check of nproc_ortho seems to be irrelevant in the
> > subroutine relating to tdosinboxes (projwfc_box.f90).
> >
> >
> > Do I misunderstand something?
> >
> >
> > Regards.
> > _______________________________________________
> > Pw_forum mailing list
> > Pw_forum at pwscf.org
> > http://pwscf.org/mailman/listinfo/pw_forum
>
> --
>  Paolo Giannozzi, Dept. Chemistry&Physics&Environment,
>  Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
>  Phone +39-0432-558216, fax +39-0432-558222
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://pwscf.org/mailman/listinfo/pw_forum
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20150313/90a69acd/attachment.html>


More information about the users mailing list