[MITgcm-support] differentiating code_oad with ecco and cal enabled + OpenAd

[Suneet Dwivedi]

Hi Patrick,
Thanks for your reply. Any ideas when these augmentations will be complete?
Could you please point me to the places in the code where development is
generally needed when a new package is enabled for use with OpenAD?
Thanks again,
Suneet


On Wed, Aug 7, 2013 at 5:02 PM, Patrick Heimbach <heimbach at mit.edu> wrote:

> Hi Suneet,
>
> we haven't pushed pkg/ecco (nor pkg/cal) through OpenAD yet
> (but we've been pushing more and more packages over the last 6 months,
> and hope to get to these ones soon).
> There are likely more problems than the ones you've encountered,
> and some more code changes needed to fully sort this out.
>
> -Patrick
>
> On Aug 7, 2013, at 4:20 PM, Suneet Dwivedi <suneetdwivedi at gmail.com>
> wrote:
>
> > Dear All,
> >
> > Using the openad, I am able to build the adjoint of the example code
> given in /verification/OpenAD/code_oad. But when I try to include the
> packages ecco and cal in my model setup, I get the errors like this while
> compiling 'externalDummies_cb2m_oad.f'
> >
> >
> ------------------------------------------------------------------------------------
> > Cannot disable Fortran error message 6535
> > externalDummies_cb2m_oad.f(344): error #6535: This variable or component
> must be of a derived or structure type   [ARRAY]
> >          array_p = array%v
> > -------------------^
> >
> ---------------------------------------------------------------------------
> > Similar errors are found for many other variables like, CG2D_B, CG2D_X
> etc in the same file.
> >
> > On investigating it further, I found that the openad is not able to
> convert the 'real*8' variables like 'ARRAY', SUMPHI, CG2D_B, CG2D_X etc.
> into 'type(active)' while differentiating. Instead it converts them into
> REAL(w2f__8) and this seems to be the main cause of error.
> >
> > When I hardcode the variables 'ARRAY', SUMPHI, CG2D_B, CG2D_X into
> type(active), the externalDummies_cb2m_oad.f compiles fine.
> >
> > When ecco package is not enabled, no such errors are seen. The problem
> occurs only when 'ecco' is enabled alongwith 'cal'.
> >
> > Please help me resolve this problem. Kindly let me know what changes
> should I make in the code to convert 'real*8' variables into 'type(active)'
> as a result of differentiation using openad.
> >
> > Looking forward for your kind reply,
> >
> > Thanks,
> >
> > Suneet
>
>
> ---
> Patrick Heimbach | heimbach at mit.edu | http://www.mit.edu/~heimbach
> MIT | EAPS 54-1420 | 77 Massachusetts Ave | Cambridge MA 02139 USA
> FON +1-617-253-5259 | FAX +1-617-253-4464 | SKYPE patrick.heimbach
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mitgcm.org/pipermail/mitgcm-support/attachments/20130808/54dfefb6/attachment.htm>