Refactor barotropic time stepping and OBC code

[Hallberg-NOAA]

This series of 9 commits extensively refactors the code in MOM_barotropic for simplicity and efficiency, with a particular emphasis on reducing the cost of the open boundary condition code and using simpler subsidiary routines within the barotropic time stepping loops.

The open boundary condition code that is called from inside of btstep_timeloop() no longer uses the indirectly referenced segments in the OBC type to determine the location, direction and kind of open boundary conditions, instead using two new arrays (u_OBC_type and v_OBC_type) in the local BT_OBC_type to specify the kind of OBC point with their absolute value while their signs indicate the direction. The OBC code in the barotropic timestepping is only called now if there is an OBC segment on the PE, and the loop ranges over which the OBCs are applied are closely tailored to cover the ranges over which eastern, western, northern or southern boundary conditions are applied on a PE. These restricted loop ranges take advantage of the fact that often there are only open boundary condition points at the edges of the domain, while still allowing for the more general case where open boundary condition segments can be anywhere. In other places, the multiplication of BT_force_[uv] by the new OBC masking arrays OBCmask_[uv] and a reset of bt_rem_[uv] during the setup phase eliminate the need for separate OBC-specific loops during the barotropic time stepping. Also revised btcalc() to usethe BT_OBC_type to streamline the application of open boundary conditions in the calculation of frhatu and frhatv. All of the extra logic related to the OBCs can be determined during initialization (specifically in the new routine initialize_BT_OBC()), so it should have a negligible impact on runtime, whereas the restructuring of the OBC code should reduce its computational burden during the time stepping itself.

Other changes are separate from the updates to the OBC code, with a rearrangement of the code that makes up btstep_timeloop(). The pressure gradient force calculation is now in the new subroutine btloop_find_PF(), which allows for the elimination of the eta_PF_BT pointer and a substantial reduction in the number of arguments to btloop_update_u() and btloop_update_v(). The 8 separate pseudo-Coriolis arrays ([abcd]zon and [abcd]mer) were combined into the new 3-d arrays f_4_u and f_4_v, with the new 4-point fastest varying index covering the influence of the 4 nearest v-velocity anomalies on the Coriolis acceleration at a u-point. The scope of btloop_eta_predictor() was reduced to just the predicting eta, with the calculation of p_surf_dyn, eta_sum and eta_PF occurring outside or in different routines. The code to calculate the dynamic pressure was moved into the new routine btloop_add_dyn_PF().

Several spelling errors in comments were also corrected, and some comments describing variables were expanded for clarity.

All answers are bitwise identical and no public interfaces are changed. The specific commit in this PR include:

• 98feea2e3 Refactor btcalc to reduce duplicated code
• bed5ba7e2 Add btloop_add_dyn_PF and refactor OBCs in btcalc
• 5c101b467 Barotropic OBC memory and halo update cleanup
• a6997e7d6 Split western and eastern OBC loops
• 55839406f Refactor barotropic OBC code
• 495f7e1e8 Eliminate eta_PF_BT pointer in btstep_timeloop
• 3e444dd39 Simpler btloop_eta_predictor
• e162d02fc Add new subroutine btloop_find_PF
• fee587084 Eliminate OBC loops in btloop_update_u

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 38.03%. Comparing base (6eca7bb) to head (c673ce7).

Additional details and impacted files

@@             Coverage Diff              @@
##           dev/gfdl     #878      +/-   ##
============================================
+ Coverage     38.01%   38.03%   +0.02%     
============================================
  Files           299      299              
  Lines         88063    88148      +85     
  Branches      16547    16578      +31     
============================================
+ Hits          33475    33525      +50     
- Misses        48513    48531      +18     
- Partials       6075     6092      +17     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[yichengt900]

This PR has passed the CEFI regression test suite (for both NWA12 and NEP10).

[theresa-cordero]

There are many changes here. The new OBC masks seem to be a good addition and I appreciate the correction of spelling errors. Since these changes do not change answers with ot without OBCs I think they should be accepted.

[marshallward]

Static build error:

../../../../../.././/src/MOM6/src/core/MOM_barotropic.F90:5691:11:

 5691 |   allocate(CS%OBCmask_u(CS%isdw-1:CS%iedw,CS%jsdw:CS%jedw), source=1.0)
      |           1
Error: Allocate-object at (1) must be ALLOCATABLE or a POINTER
../../../../../.././/src/MOM6/src/core/MOM_barotropic.F90:5692:11:

 5692 |   allocate(CS%OBCmask_v(CS%isdw:CS%iedw,CS%jsdw-1:CS%jedw), source=1.0)
      |           1
Error: Allocate-object at (1) must be ALLOCATABLE or a POINTER
make[1]: *** [Makefile:97: MOM_barotropic.o] Error 1
make[1]: *** Waiting for unfinished jobs....

[marshallward]

I think allocatable() needs to be replaced with ALLOC_ if you want to support the static build declarations for OBCmask_[uv] (and also possibly IareaT_OBCmask?)

[Hallberg-NOAA]

This PR has passed pipeline testing at https://gitlab.gfdl.noaa.gov/ogrp/mom6ci/MOM6/-/pipelines/27196.