Clean input tap

[jm-c]

What changes does this PR introduce?

• Remove duplicated parameter files from input_tap and adjust prepare_run to use instead version from input_ad.
• Remove un-tested (and un-maintained) additional dirs in exp. OpenAD (ad_singlelayer) and in lab_sea (code_tap.[adx.]noecco/).

What is the current behaviour?

• Most (if not all) parameter files for Tapenade tests are identical to TAF corresponding tests (or OpenAD in case of OpenAD).
• few untested (and not maintained) additional src and param files ; this make maintenance of the set of experiments more complicated.

What is the new behaviour

• duplicated parameter files from */input_tap have been removed ; untested additional files also.

Does this PR introduce a breaking change?

no

Other information:

Suggested addition to tag-index

o verification:

• remove duplicated parameter files from input_tap/ dir.

[mjlosch]

Looks good. Always good to remove duplcates. Note that I did not run this, because somehow my recipe to use Tapenade on my MacBook does not work anymore (incomplete notes?), and it takes me really long to figure out how to do this.

I noted a few points to consider:

1. OpenAD/code_tap/grdchk_readparms.F is a copy of pkg/grdchk/grdchk_readparms.F and could be removed.
2. These files are removed, but is it not clear to me why in this PR. Accidental? Something to check:
   • global_with_exf/input_ad/eedata.mth,
   • isomip/input_oad/eedata.mth,
   • tutorial_global_oce_biogeo/input_oad/eedata.mth
3. Probably not related to this PR, but to be considered:
   halfpipe_streamice/code_tap/ctrl_map_gentim2d.F and halfpipe_streamice/code_tap/ctrl_map_gentim2d.F are copies of code_oad, but they have been put there because OpenAD could not handle the default code. I think Tapenade can and the experiments could be modeled after taf-versions, requiring fewer customised files.

The following are just notes, where some files could be simplified, but this really does not have anything to do with this PR:

• remove to commented lines to make it more similar to pkg/cost/cost_test.F (from 3 to 1 difference): isomip/code_[ad,oad,tap]/cost_test.F
• make more similar to pkg/cost/cost_tracer.F (add ifdef ALLOW_PTRACERS, remove unused variables thetaRef,ig, jg, blank lines, to get down to 2 differences from 8):
  tutorial_global_oce_biogeo/code_[ad,oad,tap]/cost_tracer.F

[jm-c]

@mjlosch Thanks for all these good comments. In order:
(1) will remove this duplicated grdchk_readparms.F
(2) the eedata.mth is only used by testreport for multi-threaded tests (also served to decide which test "testreport -mth" will run). Since none of the TAF adjoint tests can run with more than 1 thread, these 3 eedata.mth are, at best, useless ; but also a little bit misleading, leaving the impression that these 3 set-up, at some point, have been able to run with multiple threads. So it's better to remove them, I think.
(3) I think it's a good point, but would prefer to leave changes to executable outside this PR (there is already a large enough number of removed files here) and add this as a new "bullet point" in your nice list above.

[mjlosch]

LGTM

[jm-c]

I think this PR is ready to be merged and, if no one wants to add things here, will merge it relatively soon, like tomorrow.

[Shreyas911]

Hi @jm-c, sorry for the delayed reply. I think these changes look good!

Best,
Shreyas

[jm-c]

@Shreyas911 No problem here, and thanks for the feedback.
And just to repeat: This PR does not involve any changes to any MITgcm executable (mitgcmuv*) from any experiment, and no changes to any input files neither (parameter files or binary files). Just removing few unused files and using a single copy instead of multiple.

[jm-c]

Going to merge this PR (very) soon