[MITgcm-support] Saving adjoint variables via diagnostics package and Tapenade

[Lequan Chi]

Hi all,

I am running adjoint MITgcm created via Tapenade, a recently introduced
automatic differentiation engine. The following "data.diagnostics" is
adopted to save sensitivity records with the diagnostics package. However,
the model ends normally without the expected "diag_ADJtheta*" files. No
errors or related warnings appeared. I tried to save other variables (like
"ETANSQ  ") with the diagnostics package and it works for non-adjoint
variables. The diagnostic package is also used for adjoint variables in
"global_ocean.cs32x15" with Tapenade (
https://github.com/MITgcm/MITgcm/tree/master/verification/global_ocean.cs32x15).
Similar problems happened again here: the diagnostics package can only save
nonadjoint variables. The adjoint variables are missing from the outputs
without any errors or related warnings. May I ask whether the diagnostic
package is compatible with Tapenade? If not, are there alternative ways to
save variables like ADJtheta?


Cheers
Lequan


----------------------------
data.diagnostics
----------------------------

  &DIAGNOSTICS_LIST
  diag_mnc=.true.,
  fields(1:2,1) = 'ADJtheta',
  fileName(1) = 'diag_ADJtheta',
  frequency(1) = 86400.,
  &

  &DIAG_STATIS_PARMS
  &
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.mitgcm.org/pipermail/mitgcm-support/attachments/20231223/2776726a/attachment.html>