Problem with outpost
279 views
Skip to first unread message
Vladislav Ivashchenko
Oct 10, 2019, 5:07:01 AM10/10/19
to Nek5000
Hi, Neks!
Having a strange problem. Previously I have used 'outpost' subroutine in several problems, now I have a small test with the turbChannel example. Inside userchk I have:
call outpost(vx,vy,vz,pr,pr,'neg')
write(*,*) 'OUTPOST DONE'
endif
And the code freezes at this lines (when I comment them, the calculation continues without problems). Can you tell me please where to look for the cause of the problem?
Typical f000* files are writing every Nsteps without any problems too.
Best regards,
Vladislav
Fischer, Paul
Oct 10, 2019, 10:00:28 AM10/10/19
to Vladislav Ivashchenko, Nek5000
Dear Vladislav,
Does it print the outpost done stmt?
Do you have incude 'TOTAL' in the routine that is making the call?
Paul
From: nek...@googlegroups.com <nek...@googlegroups.com> on behalf of Vladislav Ivashchenko <v.ivas...@g.nsu.ru>
Sent: Thursday, October 10, 2019 4:07 AM
To: Nek5000 <nek...@googlegroups.com>
Subject: [nek5000] Problem with outpost
Sent: Thursday, October 10, 2019 4:07 AM
To: Nek5000 <nek...@googlegroups.com>
Subject: [nek5000] Problem with outpost
--
You received this message because you are subscribed to the Google Groups "Nek5000" group.
To unsubscribe from this group and stop receiving emails from it, send an email to nek5000+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/nek5000/5fbfa9c6-ba36-4160-8886-1db802d3892e%40googlegroups.com.
You received this message because you are subscribed to the Google Groups "Nek5000" group.
To unsubscribe from this group and stop receiving emails from it, send an email to nek5000+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/nek5000/5fbfa9c6-ba36-4160-8886-1db802d3892e%40googlegroups.com.
YuHsiang Lan
Oct 10, 2019, 10:12:37 AM10/10/19
to Nek5000
Hi Vladislav,
My guess is, you are using either PNPN-2 or having multiple passive scalars, which will have lx2=lx1-1 or ldimt>1, respectively.
The usual way to call outpost is
call outpost(vx,vy,vz,pr,t,'str')
pr is of size pr(lx2,ly2,lz2,lelv) and t is of size t(lx1,ly1,lz1,lelt,ldimt)
If you want to replace any of those, the size must be the same, or larger.
Yu-Hsiang
--
On Thursday, October 10, 2019 at 9:00:28 AM UTC-5, fischerp wrote:
Dear Vladislav,
Does it print the outpost done stmt?
Do you have incude 'TOTAL' in the routine that is making the call?
Paul
From: nek...@googlegroups.com <nek...@googlegroups.com> on behalf of Vladislav Ivashchenko <v.ivas...@g.nsu.ru>
Sent: Thursday, October 10, 2019 4:07 AM
To: Nek5000 <nek...@googlegroups.com>
Subject: [nek5000] Problem with outpost
--Hi, Neks!
Having a strange problem. Previously I have used 'outpost' subroutine in several problems, now I have a small test with the turbChannel example. Inside userchk I have:
call outpost(vx,vy,vz,pr,pr,'neg')
write(*,*) 'OUTPOST DONE'
endif
And the code freezes at this lines (when I comment them, the calculation continues without problems). Can you tell me please where to look for the cause of the problem?Typical f000* files are writing every Nsteps without any problems too.
Best regards,Vladislav
You received this message because you are subscribed to the Google Groups "Nek5000" group.
To unsubscribe from this group and stop receiving emails from it, send an email to nek5000+unsubscribe@googlegroups.com.
Vladislav Ivashchenko
Oct 11, 2019, 2:25:44 AM10/11/19
to Nek5000
Dear Paul,
Thank you for your answer!
No, it doesn't print anything, just freezes, all processes are loaded (according to 'htop' command).
Yes, I have included 'TOTAL'
четверг, 10 октября 2019 г., 21:00:28 UTC+7 пользователь fischerp написал:
Dear Vladislav,
Does it print the outpost done stmt?
Do you have incude 'TOTAL' in the routine that is making the call?
Paul
From: nek...@googlegroups.com <nek...@googlegroups.com> on behalf of Vladislav Ivashchenko <v.ivas...@g.nsu.ru>
Sent: Thursday, October 10, 2019 4:07 AM
To: Nek5000 <nek...@googlegroups.com>
Subject: [nek5000] Problem with outpost
--Hi, Neks!
Having a strange problem. Previously I have used 'outpost' subroutine in several problems, now I have a small test with the turbChannel example. Inside userchk I have:
call outpost(vx,vy,vz,pr,pr,'neg')
write(*,*) 'OUTPOST DONE'
endif
And the code freezes at this lines (when I comment them, the calculation continues without problems). Can you tell me please where to look for the cause of the problem?Typical f000* files are writing every Nsteps without any problems too.
Best regards,Vladislav
You received this message because you are subscribed to the Google Groups "Nek5000" group.
To unsubscribe from this group and stop receiving emails from it, send an email to nek5000+unsubscribe@googlegroups.com.
Vladislav Ivashchenko
Oct 11, 2019, 2:29:01 AM10/11/19
to Nek5000
Dear Yu-Hsiang,
Thank you for your answer!
No, I use Pn-Pn, in SIZE file:
parameter (lx2=lx1-0) ! p-order for pressure (lx1 or lx1-2)
So, lx2=lx1 and pr has the same size like vx, vy or vz. That is why I think that outpost(vx,vy,vz,pr,pr,'neg') in my case should work fine.
Stefan Kerkemeier
Oct 11, 2019, 3:30:54 AM10/11/19
to Vladislav Ivashchenko, Nek5000
Do all MPI ranks call outpost?
To unsubscribe from this group and stop receiving emails from it, send an email to nek5000+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/nek5000/cc323853-29de-4fbe-9b18-a222e5cd9946%40googlegroups.com.
Vladislav Ivashchenko
Oct 11, 2019, 6:59:11 AM10/11/19
to Nek5000
It seems that it's not necessary to specify MPI rank for outpost in userchk subroutine, is it?
пятница, 11 октября 2019 г., 14:30:54 UTC+7 пользователь StefanK написал:
To view this discussion on the web visit https://groups.google.com/d/msgid/nek5000/cc323853-29de-4fbe-9b18-a222e5cd9946%40googlegroups.com.
Philipp Schlatter
Oct 11, 2019, 7:46:06 AM10/11/19
to nek...@googlegroups.com
ok, so it seems that your if-clause around the outpost is not giving the
same condition for all ranks at the same time. So you should probably
synchronize the count variable before the if-clause (with some proper
reduction operation). Right now count is only set to istep on those
ranks where you found negative velocities, but not necessarily on all.
This explains why outpost hangs.
Philipp
On 2019-10-11 12:59, Vladislav Ivashchenko wrote:
> It seems that it's not necessary to specify MPI rank for outpost in
> userchk subroutine, is it?
>
>
>
> пятница, 11 октября 2019 г., 14:30:54 UTC+7 пользователь StefanK написал:
>
> Do all MPI ranks call outpost?
>
> On 11 Oct 2019, at 09:24, Vladislav Ivashchenko
>> *From:* nek...@googlegroups.com
>> <mailto:nek...@googlegroups.com> <nek...@googlegroups.com
>> <mailto:nek...@googlegroups.com>> on behalf of Vladislav
>> Ivashchenko <v.ivas...@g.nsu.ru
>> <mailto:v.ivas...@g.nsu.ru>>
>> *Sent:* Thursday, October 10, 2019 4:07 AM
>> *To:* Nek5000 <nek...@googlegroups.com
>> <mailto:nek...@googlegroups.com>>
>> *Subject:* [nek5000] Problem with outpost
>> <mailto:nek5000+u...@googlegroups.com>.
>> <mailto:nek5000+u...@googlegroups.com>.
>> <https://groups.google.com/d/msgid/nek5000/cc323853-29de-4fbe-9b18-a222e5cd9946%40googlegroups.com?utm_medium=email&utm_source=footer>.
> <mailto:nek5000+u...@googlegroups.com>.
> <https://groups.google.com/d/msgid/nek5000/cd116e7d-c2de-4410-b5bb-b3f9b5123306%40googlegroups.com?utm_medium=email&utm_source=footer>.
same condition for all ranks at the same time. So you should probably
synchronize the count variable before the if-clause (with some proper
reduction operation). Right now count is only set to istep on those
ranks where you found negative velocities, but not necessarily on all.
This explains why outpost hangs.
Philipp
On 2019-10-11 12:59, Vladislav Ivashchenko wrote:
> It seems that it's not necessary to specify MPI rank for outpost in
> userchk subroutine, is it?
>
>
>
> пятница, 11 октября 2019 г., 14:30:54 UTC+7 пользователь StefanK написал:
>
> Do all MPI ranks call outpost?
>
> On 11 Oct 2019, at 09:24, Vladislav Ivashchenko
> <v.ivas...@g.nsu.ru <mailto:v.ivas...@g.nsu.ru>> wrote:
>
>> Dear Paul,
>>
>> Thank you for your answer!
>>
>> No, it doesn't print anything, just freezes, all processes are
>> loaded (according to 'htop' command).
>>
>> Yes, I have included 'TOTAL'
>>
>> четверг, 10 октября 2019 г., 21:00:28 UTC+7 пользователь fischerp
>> написал:
>>
>>
>> Dear Vladislav,
>>
>> Does it print the outpost done stmt?
>>
>> Do you have incude 'TOTAL' in the routine that is making the call?
>>
>> Paul
>>
>> ------------------------------------------------------------------------
>
>> Dear Paul,
>>
>> Thank you for your answer!
>>
>> No, it doesn't print anything, just freezes, all processes are
>> loaded (according to 'htop' command).
>>
>> Yes, I have included 'TOTAL'
>>
>> четверг, 10 октября 2019 г., 21:00:28 UTC+7 пользователь fischerp
>> написал:
>>
>>
>> Dear Vladislav,
>>
>> Does it print the outpost done stmt?
>>
>> Do you have incude 'TOTAL' in the routine that is making the call?
>>
>> Paul
>>
>> *From:* nek...@googlegroups.com
>> <mailto:nek...@googlegroups.com> <nek...@googlegroups.com
>> <mailto:nek...@googlegroups.com>> on behalf of Vladislav
>> Ivashchenko <v.ivas...@g.nsu.ru
>> <mailto:v.ivas...@g.nsu.ru>>
>> *Sent:* Thursday, October 10, 2019 4:07 AM
>> *To:* Nek5000 <nek...@googlegroups.com
>> <mailto:nek...@googlegroups.com>>
>> *Subject:* [nek5000] Problem with outpost
>> Hi, Neks!
>>
>> Having a strange problem. Previously I have used 'outpost'
>> subroutine in several problems, now I have a small test with
>> the turbChannel example. Inside userchk I have:
>>
>> |
>> if(count.eq.istep)then
>> call outpost(vx,vy,vz,pr,pr,'neg')
>> write(*,*)'OUTPOST DONE'
>> endif
>> |
>>
>>
>> And the code freezes at this lines (when I comment them, the
>> calculation continues without problems). Can you tell me
>> please where to look for the cause of the problem?
>> Typical f000* files are writing every Nsteps without any
>> problems too.
>>
>> Best regards,
>> Vladislav
>>
>> --
>> You received this message because you are subscribed to the
>> Google Groups "Nek5000" group.
>> To unsubscribe from this group and stop receiving emails from
>> it, send an email to nek5000+u...@googlegroups.com
>>
>> Having a strange problem. Previously I have used 'outpost'
>> subroutine in several problems, now I have a small test with
>> the turbChannel example. Inside userchk I have:
>>
>> |
>> if(count.eq.istep)then
>> call outpost(vx,vy,vz,pr,pr,'neg')
>> write(*,*)'OUTPOST DONE'
>> endif
>> |
>>
>>
>> And the code freezes at this lines (when I comment them, the
>> calculation continues without problems). Can you tell me
>> please where to look for the cause of the problem?
>> Typical f000* files are writing every Nsteps without any
>> problems too.
>>
>> Best regards,
>> Vladislav
>>
>> --
>> You received this message because you are subscribed to the
>> Google Groups "Nek5000" group.
>> To unsubscribe from this group and stop receiving emails from
>> <mailto:nek5000+u...@googlegroups.com>.
>> To view this discussion on the web visit
>> https://groups.google.com/d/msgid/nek5000/5fbfa9c6-ba36-4160-8886-1db802d3892e%40googlegroups.com
>> <https://groups.google.com/d/msgid/nek5000/5fbfa9c6-ba36-4160-8886-1db802d3892e%40googlegroups.com?utm_medium=email&utm_source=footer>.
>> https://groups.google.com/d/msgid/nek5000/5fbfa9c6-ba36-4160-8886-1db802d3892e%40googlegroups.com
>>
>> --
>> You received this message because you are subscribed to the Google
>> Groups "Nek5000" group.
>> To unsubscribe from this group and stop receiving emails from it,
>> send an email to nek5000+u...@googlegroups.com
>> --
>> You received this message because you are subscribed to the Google
>> Groups "Nek5000" group.
>> To unsubscribe from this group and stop receiving emails from it,
>> <mailto:nek5000+u...@googlegroups.com>.
>> To view this discussion on the web visit
>> https://groups.google.com/d/msgid/nek5000/cc323853-29de-4fbe-9b18-a222e5cd9946%40googlegroups.com
>> <https://groups.google.com/d/msgid/nek5000/cc323853-29de-4fbe-9b18-a222e5cd9946%40googlegroups.com?utm_medium=email&utm_source=footer>.
>
> --
> You received this message because you are subscribed to the Google
> Groups "Nek5000" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to nek5000+u...@googlegroups.com
> --
> You received this message because you are subscribed to the Google
> Groups "Nek5000" group.
> To unsubscribe from this group and stop receiving emails from it, send
> <mailto:nek5000+u...@googlegroups.com>.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/nek5000/cd116e7d-c2de-4410-b5bb-b3f9b5123306%40googlegroups.com
> <https://groups.google.com/d/msgid/nek5000/cd116e7d-c2de-4410-b5bb-b3f9b5123306%40googlegroups.com?utm_medium=email&utm_source=footer>.
Vladislav Ivashchenko
Oct 14, 2019, 2:41:17 AM10/14/19
to Nek5000
Thank you very much!
I've just added the 'iglsum' function for the count and now it is the same for all ranks!
Best regards,
Vladislav
пятница, 11 октября 2019 г., 18:46:06 UTC+7 пользователь Philipp Schlatter написал:
>> <mailto:nek5000@googlegroups.com> <nek...@googlegroups.com
>> <mailto:nek5000@googlegroups.com>> on behalf of Vladislav
>> Ivashchenko <v.ivas...@g.nsu.ru
>> <mailto:v.ivas...@g.nsu.ru>>
>> *Sent:* Thursday, October 10, 2019 4:07 AM
>> *To:* Nek5000 <nek...@googlegroups.com
>> <mailto:nek5000@googlegroups.com>>
>> *Subject:* [nek5000] Problem with outpost
>> Hi, Neks!
>>
>> Having a strange problem. Previously I have used 'outpost'
>> subroutine in several problems, now I have a small test with
>> the turbChannel example. Inside userchk I have:
>>
>> |
>> if(count.eq.istep)then
>> call outpost(vx,vy,vz,pr,pr,'neg')
>> write(*,*)'OUTPOST DONE'
>> endif
>> |
>>
>>
>> And the code freezes at this lines (when I comment them, the
>> calculation continues without problems). Can you tell me
>> please where to look for the cause of the problem?
>> Typical f000* files are writing every Nsteps without any
>> problems too.
>>
>> Best regards,
>> Vladislav
>>
>> --
>> You received this message because you are subscribed to the
>> Google Groups "Nek5000" group.
>> To unsubscribe from this group and stop receiving emails from
>> it, send an email to nek5000+unsubscribe@googlegroups.com
>> <mailto:nek5000+unsubscribe@googlegroups.com>.
>> To view this discussion on the web visit
>> https://groups.google.com/d/msgid/nek5000/5fbfa9c6-ba36-4160-8886-1db802d3892e%40googlegroups.com
>> <https://groups.google.com/d/msgid/nek5000/5fbfa9c6-ba36-4160-8886-1db802d3892e%40googlegroups.com?utm_medium=email&utm_source=footer>.
>>
>> --
>> You received this message because you are subscribed to the Google
>> Groups "Nek5000" group.
>> To unsubscribe from this group and stop receiving emails from it,
>> send an email to nek5000+unsubscribe@googlegroups.com
>> <mailto:nek5000+unsubscribe@googlegroups.com>.
>> To view this discussion on the web visit
>> https://groups.google.com/d/msgid/nek5000/cc323853-29de-4fbe-9b18-a222e5cd9946%40googlegroups.com
>> <https://groups.google.com/d/msgid/nek5000/cc323853-29de-4fbe-9b18-a222e5cd9946%40googlegroups.com?utm_medium=email&utm_source=footer>.
>
> --
> You received this message because you are subscribed to the Google
> Groups "Nek5000" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to nek5000+unsubscribe@googlegroups.com
> <mailto:nek5000+unsubscribe@googlegroups.com>.
Reply all
Reply to author
Forward
0 new messages