Update AB2 (mpas_ocn_time_integration_split_ab2.F) code

[hyungyukang]

This PR updates AB2 time stepping code.

Some codes and a routine were updated in split.F, but not in split_ab2.F

• ocn_diagnostic_solve_wctEdge routine was updated, but split_ab2.F was using a call sentence of the old version of the routine.
  • deltaSSH computation was missing in split_ab2.F.
  • baroclinicThickness is computed instead of 'bottomDepthEdge'.
• 'Direct application of tidal boundary condition', which was only in split.F, is added to split_ab2.F.

[github-actions]

PR Preview Action v1.4.7
🚀 Deployed preview to https://E3SM-Project.github.io/E3SM/pr-preview/pr-6323/
on branch gh-pages at 2024-04-03 17:45 UTC

[hyungyukang]

Passed all COMPASS nightly suite on Perlmutter with GNU compiler.

Test Runtimes:
00:49 PASS ocean_baroclinic_channel_10km_default
01:38 PASS ocean_baroclinic_channel_10km_threads_test
00:44 PASS ocean_baroclinic_channel_10km_decomp_test
00:27 PASS ocean_baroclinic_channel_10km_restart_test
28:03 PASS ocean_global_ocean_QU240_mesh
01:56 PASS ocean_global_ocean_QU240_WOA23_init
00:55 PASS ocean_global_ocean_QU240_WOA23_performance_test
02:30 PASS ocean_global_ocean_QU240_WOA23_restart_test
02:02 PASS ocean_global_ocean_QU240_WOA23_decomp_test
02:04 PASS ocean_global_ocean_QU240_WOA23_threads_test
01:37 PASS ocean_global_ocean_QU240_WOA23_analysis_test
01:13 PASS ocean_global_ocean_QU240_WOA23_RK4_performance_test
01:59 PASS ocean_global_ocean_QU240_WOA23_RK4_restart_test
01:55 PASS ocean_global_ocean_QU240_WOA23_RK4_decomp_test
03:12 PASS ocean_global_ocean_QU240_WOA23_RK4_threads_test
00:00 PASS ocean_global_ocean_QUwISC240_mesh
00:00 PASS ocean_global_ocean_QUwISC240_WOA23_init
03:00 PASS ocean_global_ocean_QUwISC240_WOA23_performance_test
00:41 PASS ocean_ice_shelf_2d_5km_z-star_restart_test
00:23 PASS ocean_ice_shelf_2d_5km_z-level_restart_test
00:21 PASS ocean_ziso_20km_default
00:27 PASS ocean_ziso_20km_with_frazil
Total runtime 56:00
PASS: All passed successfully!

[mark-petersen]

@hyungyukang commented in a meeting that this is the critical change in this PR.

[cbegeman]

It looks from #6047 that this would be BFB when the thickness is centered, so I don't think this will affect the instability we encountered in the v3 HR G-case.

[hyungyukang]

@cbegeman , thanks for your comment. When I tested this branch very first time, I was able to see some differences in simple test cases. So I thought this change is critical. But today, I cleaned and compiled again after the meeting, and it is very similar now. Maybe I did something wrong during early tests. However, for code uniformity between split_explicit_ab2.F and split_explicit.F, this PR is worth merging.

This PR is not BFB with very small differences (~1e-9).

Just in case, I will test this PR a bit more.

[mark-petersen]

@hyungyukang, it would be better to separate this PR into two PRs: one that is BFB on all tests with the majority of changes, and one with the remaining non-BFB lines. That way it is clear what code changes are non-BFB. It looks like many of these changes are clean-up and conforming to the split-explicit code, but are not changing the computation. It is very helpful to have non-BFB PRs with small isolated changes, even just a few lines, when we look back later and ask how it changed the results.

[hyungyukang]

@hyungyukang, it would be better to separate this PR into two PRs: one that is BFB on all tests with the majority of changes, and one with the remaining non-BFB lines. That way it is clear what code changes are non-BFB. It looks like many of these changes are clean-up and conforming to the split-explicit code, but are not changing the computation. It is very helpful to have non-BFB PRs with small isolated changes, even just a few lines, when we look back later and ask how it changed the results.

@mark-petersen , Thanks for suggesting. Agreed. I will close this PR first, then open one PR with BFB on all tests. Maybe it'd be better to open another PR after merging the BFB PR. Thanks!

[hyungyukang]

Closing this PR and splitting it into two PRs to clarify the BFB/non-BFB changes as suggested by a reviewer.