[MITgcm-support] diagnostic errors
Patrick Heimbach
heimbach at MIT.EDU
Mon Oct 17 20:10:23 EDT 2005
Hi Matt,
what you're seeing is very likely
particular to the use of diagnostics in
conjunction with the adjoint.
Because some of the code gets replaced by
TAF-generated code, there might be code missing
that's not required to compute the cost function.
Usually I take care of this by updating corresponding
flow directives, but things in pkg/diagnostics have
been changing so frequently that it was/is impossible
to keep up.
I'll try to sort this out this week.
Cheers
-Patrick
On Sun, 2005-10-16 at 16:32, Matthew Mazloff wrote:
> Hi Andrea,
>
> Thanks for the help.
>
> So changes that occurrred possibly causing this error are:
> 1) The model update (previous version was from Aug 25 2005) on Sep 27, 2005
> 2) Platform change, I was running on a SGI Altix and now I'm running on
> an IBM SP4
> 3) I did alter the input files slightly....but I do not think I made any
> mistakes that would cause the diagnostics package not to fill the fields
> I did change from globalfiles=.TRUE., to globalfiles=.FALSE., would
> this cause this error?
>
> Other info:
> I am using useSingleCpuIO=.TRUE.
> I am not using the mnc package
>
> I have put a copy of STDOUT.0000 in /ocean/data4/mmazloff/ so please
> feel free to check that out.
>
> Thanks a lot!
> Matt
>
>
> Andrea Molod wrote:
>
> > hi matt,
> >
> >> I have used the diagnostics package many times before without
> >> problems, but my
> >> last run I recieved the errors:
> >>
> >> (PID.TID 0000.0001) WARNING DIAGNOSTICS_OUT => write ZEROS instead
> >> (PID.TID 0000.0001) - WARNING - from DIAGNOSTICS_OUT at iter= 1460
> >> (PID.TID 0000.0001) - WARNING - diag.# 23 : ETAN (# 1 ) in
> >> outp.Stream:
> >> ETAN_bar
> >> (PID.TID 0000.0001) - WARNING - has not been filled (ndiag= 0 )
> >> (PID.TID 0000.0001) WARNING DIAGNOSTICS_OUT => write ZEROS instead
> >> (PID.TID 0000.0001) - WARNING - from DIAGNOSTICS_OUT at iter= 1460
> >> (PID.TID 0000.0001) - WARNING - diag.# 30 : UVEL (# 1 ) in
> >> outp.Stream:
> >> UVvelbar
> >> (PID.TID 0000.0001) - WARNING - has not been filled (ndiag= 0 )
> >> (PID.TID 0000.0001) WARNING DIAGNOSTICS_OUT => write ZEROS instead
> >> (PID.TID 0000.0001) - WARNING - from DIAGNOSTICS_OUT at iter= 1460
> >> (PID.TID 0000.0001) - WARNING - diag.# 31 : VVEL (# 2 ) in
> >> outp.Stream:
> >> UVvelbar
> >> (PID.TID 0000.0001) - WARNING - has not been filled (ndiag= 0 )
> >> .....etc....
> >
> >
> > this error message is telling you that the diagnostic is 'turned on'
> > properly, that is, the model has read your data.diagnostics and
> > determined
> > that there is enough space, etc..., but that the call to fill it never
> > happened.
> >
> >
> >> I have changed a few things in my setup, but cannot determine where
> >> this error
> >> arises from. Could someone please tell me how this may have happened...
> >
> >
> > hard (impossible for me) to even begin to think about what is the
> > problem unless you let us know what your experiment config is and what
> > things you have changed in your setup. even better - can you tell us
> > where to find the STDOUT(s) from your run to see what's happening there?
> >
> >> and in data.diagnostics I have
> >>
> >> &diagnostics_list
> >> #---
> >> frequency(1) = 2628000.0,
> >> levels(1,1) = 1.,
> >> fields(1,1) = 'ETAN ',
> >> filename(1) = 'ETAN_bar',
> >> #---
> >> frequency(2) = 2628000.0,
> >> fields(1,2) = 'UVEL ','VVEL ',
> >> filename(2) = 'UVvelbar',
> >> #---
> >
> >
> >
> > this seems ok to me.
> >
> > one more peice of info, if you could... how long ago did you 'update'
> > your
> > code and recompile to make this run?
> >
> > thanks, and hope i can be more help with more info.
> >
> > andrea
>
>
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support
--
--------------------------------------------------------
Patrick Heimbach Massachusetts Institute of Technology
FON: +1/617/253-5259 EAPS, Room 54-1518
FAX: +1/617/253-4464 77 Massachusetts Avenue
mailto:heimbach at mit.edu Cambridge MA 02139
http://www.mit.edu/~heimbach/ USA
More information about the MITgcm-support
mailing list