[RegCNET] بخصوص: بØ(R)صوص: CBMZ chemistry crash
James Ciarlo`
james.ciarlo at physics.org
Sat Aug 22 15:56:49 CEST 2015
Hi Zeinab,
Yes that is the variable that was suggested however, for me, the original
error showed up for dtchesolv 900, 600, 300, 150, 120 and 90.
When I tried 60 and 30 I obtained a different crash (below).
forrtl: severe (71): integer divide by zero
Image PC Routine Line Source
regcmMPI 0000000000F492C9 Unknown Unknown Unknown
regcmMPI 0000000000F47B9E Unknown Unknown Unknown
regcmMPI 0000000000EFBAD2 Unknown Unknown Unknown
regcmMPI 0000000000E98593 Unknown Unknown Unknown
regcmMPI 0000000000E9EC51 Unknown Unknown Unknown
libpthread.so.0 00002AAAAC96EBE0 Unknown Unknown Unknown
regcmMPI 00000000008963E9 Unknown Unknown Unknown
regcmMPI 000000000070B7A8 Unknown Unknown Unknown
regcmMPI 00000000004613D4 Unknown Unknown Unknown
regcmMPI 0000000000784103 Unknown Unknown Unknown
regcmMPI 00000000004611E6 Unknown Unknown Unknown
libc.so.6 00002AAAACB99994 Unknown Unknown Unknown
regcmMPI 00000000004610D9 Unknown Unknown Unknown
It seems that decreasing dtchesolv does not solve the issue, but only
creates a further complication.
I still haven't managed to solve this crash. Let me know if you come up
with anything.
Regards,
James
On 22 August 2015 at 12:04, <zeinab.salah at yahoo.com> wrote:
> Dear All,
>
> There is a variable of "dtchsolv" in mod_che_chemistry.f90, do you
> mean that?
>
> in this module :
> real(rk8) , parameter :: dtchsolv = 900.0D0
>
> if dt of my run = 150 s , should I modify dtchsolv to 150 s?
>
> Best regards,
> Zeinab
> ------------------------------
> *من:* James Ciarlo` <james.ciarlo at physics.org>
> *إلى:* Solmon Fabien <fsolmon at ictp.it>
> *نسخة كربونية:* zeinab.salah at yahoo.com; RegCNET <regcnet at lists.ictp.it>; "
> ashalaby at ictp.it" <ashalaby at ictp.it>
> *تاريخ الإرسال:* السبت 15 أغسطس، 2015 10:50 ص
> *الموضوع:* Re: [RegCNET] بØ(R)صوص: CBMZ chemistry crash
>
> Hi again everyone,
>
> I have tried reducing the dt stepwise (since I could not find dtchem) and
> have so far brought it down to 30. I am afraid that the model is still
> crashing in the same manner.
>
> I am currently scouring the code to try to understand what may be causing
> this error, as it has now also popped up in my CB6 simulation (not just
> CBMZ). I am attaching the error again for reference.
>
> Thanks and regards,
> James
>
>
>
> On 2 August 2015 at 14:18, James Ciarlo` <james.ciarlo at physics.org> wrote:
>
> Hi again,
>
> I have tried to use the dtchem but I could not find it anywhere in the
> User Guide or the model code, even after doing an svn update.
>
> On 2 August 2015 at 12:11, James Ciarlo` <james.ciarlo at physics.org> wrote:
>
> Yes, it's just what I feared. I'll try doing just that.
>
> On 1 August 2015 at 15:35, Solmon Fabien <fsolmon at ictp.it> wrote:
>
> Hello James
>
> if you do not save chemall all the fast (not transported) species are
> reset to zero , the solver will recalculate a new equilibrium at every
> time step .. in fact with this fix you might still generate unrealistic
> concentrations for some untransported species but since you reset to zero
> there are no further consequence and the code is stable , however it does
> not mean that it performs a good chemistry ... I would try to screen the
> to the solver inputs just before it starts exploding, and try to play with
> chemistry step dtchem.. try to reduce it from 900 s to dt maybe ..
>
> Fabien
> > Hi everyone,
> >
> > I have recently tried commenting the chemall sections in the
> > mod_che_chemistry.F90, as suggested by Ahmed.
> >
> >
> > ! 1 : initialise xrin with the concentrations from
> > ! previous chemsolv step
> > ! do ic = 1 , totsp
> > ! xrin(ic) = chemall(j,i,k,ic)
> > ! end do
> >
> > and
> >
> > ! save the concentrations of all species for next chemistry
> step
> > ! do ic = 1 , totsp
> > ! chemall(j,i,k,ic) = xrout(ic)
> > ! end do
> > ! Store photolysis rates for diagnostic
> >
> > Dear Ahmed,
> >
> > The simulation has just run to completion this way. however, I am
> > concerned
> > that this way, there is no chemistry continuity between time steps. It
> > this
> > the case? if so, how can we resolve this?
> >
> > Best regards,
> > James
> >
> >
> >
> > On 1 August 2015 at 12:49, <zeinab.salah at yahoo.com> wrote:
> >
> >> Hi James,
> >>
> >> We try to run the recent version also with CBMZ or DCCB and we get this
> >> error also.
> >> The file of regcm.in and the error message are in the attachments.
> >>
> >> Best regards,
> >> Zeinab
> >>
> >>
> >>
> >> ------------------------------
> >> *من:* James Ciarlo` <james.ciarlo at physics.org>
> >> *إلى:* RegCNET <regcnet at lists.ictp.it>
> >> *نسخة كربونية:* "ashalaby at ictp.it" <ashalaby at ictp.it>
> >> *تاريخ الإرسال:* الثلاثاء 21 يوليو، 2015
> >> 2:01 م
> >> *الموضوع:* [RegCNET] CBMZ chemistry crash
> >>
> >> Dear RegCNETers,
> >>
> >> I have installed a recent version of the RegCM and tried to run a CBMZ
> >> simulation for 1 year. However, the chemistry crashed and the simulation
> >> kept on running producing a massive error log.
> >>
> >> After including a few lines in the mod_che_chemistry.F90 I got the
> >> following information on when the error is happening:
> >>
> >> xrout, xrin, trac%indcbmz
> >> 7.864835401599373E+153 38524086126.5782 10
> >>
> >> 10 is H2O2.
> >>
> >> I have attached the regcm.in file and the error obtained. I cannot
> >> figure
> >> out why this is happening. I was wondering if anyone has any idea on the
> >> matter.
> >>
> >> Thanks and regards,
> >> James
> >>
> >> _______________________________________________
> >> RegCNET mailing list
> >> RegCNET at lists.ictp.it
> >> https://lists.ictp.it/mailman/listinfo/regcnet
> >>
> >>
> > _______________________________________________
> > RegCNET mailing list
> > RegCNET at lists.ictp.it
> > https://lists.ictp.it/mailman/listinfo/regcnet
>
>
> --
> Fabien Solmon
> Research Scientist
> +39 040 2240206 (office) | Earth System Physics
> +39 040 2240449 (fax) | ICTP (UNESCO-IAEA)
> esp.ictp.it | Strada Costiera 11
> | 34014 Trieste, Italy
>
>
>
>
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ictp.it/pipermail/regcnet/attachments/20150822/12dc1a08/attachment.html>
More information about the RegCNET
mailing list