[Pw_forum] problems with pp.x for parallel processing
On Oct 20, 2011, at 15:48 , Natalia Pavlenko wrote: > thank you for your help with pp.x, after removing the > lines in postproc.f90 mentioned in your last message, > I have recomplied it and now it works well without > k-parallelization and with wf_collect=.true. in the scf-input. ...and I am no longer sure what the preblem was, because I have tried again to fix it and cannot reproduce it any longer...P. --- Paolo Giannozzi, Dept of Chemistry, Univ. Udine, via delle Scienze 208, 33100 Udine, Italy Phone +39-0432-558216, fax +39-0432-558222
[Pw_forum] problems with pp.x for parallel processing
Dear Paolo,
thank you for your help with pp.x, after removing the lines in postproc.f90
mentioned in your last message, I have recomplied it and now it works well
without k-parallelization and with wf_collect=.true. in the scf-input.
Best regards, N.Pavlenko
---
Dr. Natalia Pavlenko
Institute for Condensed Matter Physics,
Lviv, Ukraine
Tel.: +38-0322-707401
Fax: +38-0322-761158
--- On Mon, 10/10/11, Paolo Giannozzi wrote:
> From: Paolo Giannozzi
> Subject: Re: [Pw_forum] problems with pp.x for parallel processing
> To: "PWSCF Forum"
> Date: Monday, October 10, 2011, 12:34 AM
> On Sat, 2011-10-08 at 22:54 -0700,
> Natalia Pavlenko wrote:
>
> > I know that pp.x works well without k-parallelization
>
> I had understood the opposite from one of your previous
> messages
>
> >? but for this I need to recalculate nscf with
> wf_collect
> > which takes pretty much time and more memory sources.
>
> if you run the nscf calcul;ation with npool*nproc_pool
> processors,
> you may actually run the postprocessing on nproc_pool
> processors
> with npool=1. You have to remove the following lines from
> postproc.f90:
> IF (nproc /= nproc_file .and. .not. twfcollect .and.
> needwf)? &
> ? ???CALL errore('postproc',&
> ? ???'pw.x run with a different number
> of processors. Use
> wf_collect=.true.',1)
>
> P.
> --
> Paolo Giannozzi, IOM-Democritos and University of Udine,
> Italy
>
>
> ___
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum
>
[Pw_forum] problems with pp.x for parallel processing
On Sat, 2011-10-08 at 22:54 -0700, Natalia Pavlenko wrote:
> I know that pp.x works well without k-parallelization
I had understood the opposite from one of your previous
messages
> but for this I need to recalculate nscf with wf_collect
> which takes pretty much time and more memory sources.
if you run the nscf calcul;ation with npool*nproc_pool processors,
you may actually run the postprocessing on nproc_pool processors
with npool=1. You have to remove the following lines from postproc.f90:
IF (nproc /= nproc_file .and. .not. twfcollect .and. needwf) &
CALL errore('postproc',&
'pw.x run with a different number of processors. Use
wf_collect=.true.',1)
P.
--
Paolo Giannozzi, IOM-Democritos and University of Udine, Italy
[Pw_forum] problems with pp.x for parallel processing
I know that pp.x works well without k-parallelization, but for this I need to recalculate nscf with wf_collect which takes pretty much time and more memory sources. Best, Natalia Dr. Natalia Pavlenko Institute for Condensed Matter Physics, Lviv, Ukraine Tel.: +38-0322-707401 Fax: +38-0322-761158 --- On Sat, 10/8/11, Paolo Giannozzi wrote: > From: Paolo Giannozzi > Subject: Re: [Pw_forum] problems with pp.x for parallel processing > To: "PWSCF Forum" > Date: Saturday, October 8, 2011, 1:07 PM > There is definitely some problem with > k-point parallelization > (-npool option) in pp.x. However, it worked for me if the > "wf_collect" option is used and pp.x is launched without > k-point parallelization > > P. > --- > Paolo Giannozzi, Dept of > Chemistry, > 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://www.democritos.it/mailman/listinfo/pw_forum >
[Pw_forum] problems with pp.x for parallel processing
There is definitely some problem with k-point parallelization (-npool option) in pp.x. However, it worked for me if the "wf_collect" option is used and pp.x is launched without k-point parallelization P. --- Paolo Giannozzi, Dept of Chemistry, Univ. Udine, via delle Scienze 208, 33100 Udine, Italy Phone +39-0432-558216, fax +39-0432-558222
[Pw_forum] problems with pp.x for parallel processing
In my opinion, the easiest way to solve this problem is to add wf_collect=.t.
in nscf calculation "input" file, then use pp.x output without
parallelism.
--
GAO Zhe
CMC Lab, MSE, SNU, Seoul, S.Korea
At 2011-10-08 02:09:04,"Natalia Pavlenko" wrote:
>I tried two different possibilities: first, I used the same number of pools
>("-npool 2") in nscf and in pp.x. In this case I received the message:
>-
>from local_dos : error # 1
> problems with xk_pool
>--
>Second, I tried to run pp.x without setting the option npool
>(just pp.x < case_rho.in > case_rho.out), in this case I get the following
>message after the crash:
>--
>from postproc : error # 1
>pw.x run with a different number of pools. Use wf_collect=.true.
>---
>
>Best regards, N. Pavlenko
>Dr. Natalia Pavlenko
>Institute for Condensed Matter Physics,
>Lviv, Ukraine
>
>Tel.: +38-0322-707401
>Fax: +38-0322-761158
>
>
>
>--- On Fri, 10/7/11, Paolo Giannozzi wrote:
>
>> From: Paolo Giannozzi
>> Subject: Re: [Pw_forum] problems with pp.x for parallel processing
>> To: "PWSCF Forum"
>> Date: Friday, October 7, 2011, 10:40 AM
>>
>> On Oct 7, 2011, at 18:19 , Natalia Pavlenko wrote:
>>
>> > I am trying to calculate charge density plots using
>> pp.x for 8
>> > parallel
>> > processors. The scf and nscf calculations are
>> successfully finished.
>> > With pp.x, I receive the following warning after the
>> CRASH:
>>
>> you didn't try to run on a diferent number of pools from
>> the one
>> used in the nscf calculation?
>>
>> P.
>> ---
>> Paolo Giannozzi, Dept of
>> Chemistry,
>> 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://www.democritos.it/mailman/listinfo/pw_forum
>>
>___
>Pw_forum mailing list
>Pw_forum at pwscf.org
>http://www.democritos.it/mailman/listinfo/pw_forum
-- next part --
An HTML attachment was scrubbed...
URL:
http://www.democritos.it/pipermail/pw_forum/attachments/20111008/27dba665/attachment-0001.htm
[Pw_forum] problems with pp.x for parallel processing
Dear Paolo, In attachment I am sending 3 job scripts for a simple SrTiO3 layered structure which can be calculated pretty fast: (1)scf, (2)nscf and (3) pp.x for evaluation of charge density contours. The scf and nscf calculations are performed without any problem, but pp.x is crashed with the same warning: from local_dos : error # 1 problems with xk_pool I would be grateful for your assistance. With best regards, Natalia - Dr. Natalia Pavlenko Institute for Condensed Matter Physics, Lviv, Ukraine Tel.: +38-0322-707401 Fax: +38-0322-761158 --- On Fri, 10/7/11, Paolo Giannozzi wrote: > From: Paolo Giannozzi > Subject: Re: [Pw_forum] problems with pp.x for parallel processing > To: "PWSCF Forum" > Date: Friday, October 7, 2011, 1:32 PM > > On Oct 7, 2011, at 22:29 , Natalia Pavlenko wrote: > > > It does not work, unfortunately. > > please provide an example that can be run in a reasonable > amount of time > > P. > --- > Paolo Giannozzi, Dept of > Chemistry, > 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://www.democritos.it/mailman/listinfo/pw_forum > -- next part -- A non-text attachment was scrubbed... Name: 0job Type: application/octet-stream Size: 3353 bytes Desc: not available Url : http://www.democritos.it/pipermail/pw_forum/attachments/20111008/0f7330ed/attachment.obj -- next part -- A non-text attachment was scrubbed... Name: 0job_nscf Type: application/octet-stream Size: 3407 bytes Desc: not available Url : http://www.democritos.it/pipermail/pw_forum/attachments/20111008/0f7330ed/attachment-0001.obj -- next part -- A non-text attachment was scrubbed... Name: 0job_rho Type: application/octet-stream Size: 2500 bytes Desc: not available Url : http://www.democritos.it/pipermail/pw_forum/attachments/20111008/0f7330ed/attachment-0002.obj
[Pw_forum] problems with pp.x for parallel processing
On Oct 7, 2011, at 22:29 , Natalia Pavlenko wrote: > It does not work, unfortunately. please provide an example that can be run in a reasonable amount of time P. --- Paolo Giannozzi, Dept of Chemistry, Univ. Udine, via delle Scienze 208, 33100 Udine, Italy Phone +39-0432-558216, fax +39-0432-558222
[Pw_forum] problems with pp.x for parallel processing
On Oct 7, 2011, at 18:19 , Natalia Pavlenko wrote: > I am trying to calculate charge density plots using pp.x for 8 > parallel > processors. The scf and nscf calculations are successfully finished. > With pp.x, I receive the following warning after the CRASH: you didn't try to run on a diferent number of pools from the one used in the nscf calculation? P. --- Paolo Giannozzi, Dept of Chemistry, Univ. Udine, via delle Scienze 208, 33100 Udine, Italy Phone +39-0432-558216, fax +39-0432-558222
[Pw_forum] problems with pp.x for parallel processing
It does not work, unfortunately.
Dr. Natalia Pavlenko
Institute for Condensed Matter Physics,
Lviv, Ukraine
Tel.: +38-0322-707401
Fax: +38-0322-761158
--- On Fri, 10/7/11, GAO Zhe wrote:
From: GAO Zhe <flux_ra...@163.com>
Subject: Re: [Pw_forum] problems with pp.x for parallel processing
To: "PWSCF Forum"
Date: Friday, October 7, 2011, 12:23 PM
In my opinion, the easiest way to solve this problem is to add wf_collect=.t.
in nscf calculation "input" file, then use pp.x output without
parallelism.
--
GAO Zhe
CMC Lab, MSE, SNU, Seoul, S.Korea
At?2011-10-08?02:09:04,"Natalia?Pavlenko"??wrote:
>I?tried?two?different?possibilities:?first,?I?used?the?same?number?of?pools?("-npool?2")?in?nscf?and?in?pp.x.?In?this?case?I?received?the?message:
>-
>from?local_dos?:?error?#?1
>?problems?with?xk_pool
>--
>Second,?I?tried?to?run?pp.x?without?setting?the?option?npool
>(just?pp.x??case_rho.out),?in?this?case?I?get?the?following?message?after?the?crash:
>--
>from?postproc?:?error?#?1
>pw.x?run?with?a?different?number?of?pools.?Use?wf_collect=.true.
>---
>
>Best?regards,?N.?Pavlenko
>Dr.?Natalia?Pavlenko
>Institute?for?Condensed?Matter?Physics,
>Lviv,?Ukraine
>
>Tel.:?+38-0322-707401
>Fax:?+38-0322-761158
>
>
>
>---?On?Fri,?10/7/11,?Paolo?Giannozzi??wrote:
>
>>?From:?Paolo?Giannozzi?
>>?Subject:?Re:?[Pw_forum]?problems?with?pp.x?for?parallel?processing
>>?To:?"PWSCF?Forum"?
>>?Date:?Friday,?October?7,?2011,?10:40?AM
>>?
>>?On?Oct?7,?2011,?at?18:19?,?Natalia?Pavlenko?wrote:
>>?
>>?>?I?am?trying?to?calculate?charge?density?plots?using
>>?pp.x?for?8??
>>?>?parallel
>>?>?processors.?The?scf?and?nscf?calculations?are
>>?successfully?finished.
>>?>?With?pp.x,?I?receive?the?following?warning?after?the
>>?CRASH:
>>?
>>?you?didn't?try?to?run?on?a?diferent?number?of?pools?from
>>?the?one
>>?used?in?the?nscf?calculation?
>>?
>>?P.
>>?---
>>?Paolo?Giannozzi,?Dept?of
>>?Chemistry,
>>?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://www.democritos.it/mailman/listinfo/pw_forum
>>?
>___
>Pw_forum?mailing?list
>Pw_forum at pwscf.org
>http://www.democritos.it/mailman/listinfo/pw_forum
-Inline Attachment Follows-
___
Pw_forum mailing list
Pw_forum at pwscf.org
http://www.democritos.it/mailman/listinfo/pw_forum
-- next part --
An HTML attachment was scrubbed...
URL:
http://www.democritos.it/pipermail/pw_forum/attachments/20111007/96f061c7/attachment.htm
[Pw_forum] problems with pp.x for parallel processing
I tried two different possibilities: first, I used the same number of pools
("-npool 2") in nscf and in pp.x. In this case I received the message:
-
from local_dos : error # 1
problems with xk_pool
--
Second, I tried to run pp.x without setting the option npool
(just pp.x < case_rho.in > case_rho.out), in this case I get the following
message after the crash:
--
from postproc : error # 1
pw.x run with a different number of pools. Use wf_collect=.true.
---
Best regards, N. Pavlenko
Dr. Natalia Pavlenko
Institute for Condensed Matter Physics,
Lviv, Ukraine
Tel.: +38-0322-707401
Fax: +38-0322-761158
--- On Fri, 10/7/11, Paolo Giannozzi wrote:
> From: Paolo Giannozzi
> Subject: Re: [Pw_forum] problems with pp.x for parallel processing
> To: "PWSCF Forum"
> Date: Friday, October 7, 2011, 10:40 AM
>
> On Oct 7, 2011, at 18:19 , Natalia Pavlenko wrote:
>
> > I am trying to calculate charge density plots using
> pp.x for 8
> > parallel
> > processors. The scf and nscf calculations are
> successfully finished.
> > With pp.x, I receive the following warning after the
> CRASH:
>
> you didn't try to run on a diferent number of pools from
> the one
> used in the nscf calculation?
>
> P.
> ---
> Paolo Giannozzi, Dept of
> Chemistry,
> 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://www.democritos.it/mailman/listinfo/pw_forum
>
[Pw_forum] problems with pp.x for parallel processing
Dear users,
I am trying to calculate charge density plots using pp.x for
8 parallel processors. The scf and nscf calculations are successfully finished.
With pp.x, I receive the following warning after the CRASH:
-
task # 0
from local_dos : error # 1
problems with xk_pool
-
In my case, the variable npool=2, which is set in the job script file.
In the subroutine local_dos, I have found the following commands:
from local_dos-
IF (npool>1) THEN
CALL xk_pool( kpoint, nkstot, kpoint_pool, which_pool )
IF (kpoint_pool<1 .or. kpoint_pool> nks) &
CALL errore('local_dos','problems with xk_pool',1)
i_am_the_pool=(my_pool_id==which_pool)
ELSE
i_am_the_pool=.true.
kpoint_pool=kpoint
ENDIF
--
How to set up kpoint_pool in this case? In my structure, the k-points are
automatically defined: K_POINTS automatic
I would be grateful for any advices.
Regards,
Natalia Pavlenko
---
Dr. Natalia Pavlenko
Institute for Condensed Matter Physics,
Lviv, Ukraine
Tel.: +38-0322-707401
Fax: +38-0322-761158
