dev-master-candidate-2019-12-17 #1046
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Label output for TCs
This patch consists of three changes to the test suite: - Tests are now run in dedicated work directories. This allows for parallel execution of tests inside of make. Unexpected improvements are a 3-4x speedup on test time, and the disappearance of several anomalous "negative zero" issues which were otherwise unreproducible outside of the test, suggesting that replacing the serial runs inside of the tc directories has resolved some otherwise unknown problems. - Makefile rules for running tests are now defined relative to the .testing directory, rather than an explicit base directory based on the path of the Makefile. Although this method is slightly less robust and slightly more prone to error, it makes it easier to build and run individual files within the project. Build paths are still absolute, and will be updated in a separate commit. - Test output is now stored in a `results` directory, rather than inside the tc* configuration directories.
Test run work and results dir; relative path rules
In the discontinuous mode of neutral diffusion, the edge values need to be evaluated in the same way that the 'polynomial' evaluation returns which might be different at roundoff. This was accidentally outside of the `if CS%discontinuous` block and was moved inside
This patch addresses the inconsistency of signed zero in the minimum and maximum values used in checksum report. The behavior of the Fortran intrinsic min() and our MPI library's implementation of MPI_Reduce with MPI_MIN can give different results for different values of signed zero, e.g. min(0,-0) vs min(-0,0). Additionally, the MPI_Reduce result appears to not consistenty follow these rules in more complex MPI calculations. Due to these issues, we add the result to positive zero to ensure that any negative zero results are reported as positive.
Added a new optional scale argument to calculate_density, calculate_spec_vel calculate_density_derivs, calculate_density_second_derivs, and calculate_specific_vol_derivs, to rescale the densities or related variables. All answers are bitwise identical, but there are new optional arguments to public interfaces.
Rescale bulkmixedlayer densities and their derivatives via the calls to calculate_density and calculate_density_derivs. All answers are bitwise identical.
Rescaled density units in MOM_entrain_diffusive for dimensional consistency testing. All answers are bitwise identical.
Changed the units of GV%Rlay from [kg m-3] to [R] for dimensional consistency testing. This required the addition of unit_scale_type arguments to several interfaces. All answers are bitwise identical, but new arguments have been added to several public interfaces.
Moved rescaling of Rlay to [R] into the various set_coord routines. This required the addition of unit_scale_type arguments to two interfaces. All answers are bitwise identical, but new arguments have been added to two public interfaces.
Changed the units of GV%Rho0 from [kg m-3] to [R] for dimensional consistency testing. This required the addition of unit_scale_type arguments to several interfaces. All answers are bitwise identical, but new arguments have been added to several public interfaces and the units of an element in a public type have changed.
Rescaled density units in MOM_regularize_layers for dimensional consistency testing. All answers are bitwise identical.
Rescaled density units in MOM_set_viscosity for dimensional consistency testing. All answers are bitwise identical.
Rescaled density units in MOM_set_diffusivity for dimensional consistency testing. All answers are bitwise identical.
Rescaled density units in MOM_kappa_shear for dimensional consistency testing. All answers are bitwise identical.
Rescaled density units in MOM_internal_tide_input for dimensional consistency testing. All answers are bitwise identical.
Rescaled density units in diagnoseMLDbyDensityDifference in MOM_diabatic_aux for dimensional consistency testing. All answers are bitwise identical.
Rescaled density units in MOM_tidal_mixing for dimensional consistency testing. All answers are bitwise identical.
Rescaled density units in MOM_geothermal for dimensional consistency testing. This required adding a unit_scale_type argument to geothermal_init. All answers are bitwise identical, but a public interface has a new argument.
Corrected dimensions in comments. All answers are bitwise identical.
Partially rescaled the units of itide%TKE_itidal_input for dimensional consistency testing. All answers are bitwise identical, but the units of an element of a transparent public type have changed.
Rescaled density units in MOM_energetic_PBL for dimensional consistency testing. All answers are bitwise identical.
Rescaled the density units of the cTKE or TKE_forced variables passed to energetic_PBL and applyBoundaryFluxesInOut for dimensional consistency testing. All answers are bitwise identical, but the units of an argument to two public interfaces have changed.
Rescaled the specific volume (density) units of the dSV_dT and dSV_dS variables passed to energetic_PBL, applyBoundaryFluxesInOut, and absorbRemainingSW for dimensional consistency testing. Also rescaled the dimensions of TKE returned from absorbRemainingSW. All answers are bitwise identical, but the units of a arguments to 3 public interfaces have changed.
Used GV%H_to_RZ to simplify rescalings in applyBoundaryFluxesInOut and absorbRemainingSW. All answer are bitwise identical.
Added conversion factors to 4 mass-flux diagnostics and comments to 4 others on why no conversion factors are needed. All answers are bitwise identical.
Added scale arguments to 5 chksum calls and grouped another two chksum calls while also adding the right scaling argument. All answers are bitwise identical.
Undoes the dimensional scaling of the cell areas before taking their global sum, so that the reproducing sum does not overflow when there is dimensional rescaling. All answers are bitwise identical when there is no rescaling, but this eliminates a source of inadvertent overflows or underflows in the global sums, and there is a new optional argument to compute_global_grid_integrals.
Corrects the dimensionally inconsistent expressions for the CFL number in the tracer advection code, in which a negligible thickness had been added to the cell volume to avoid division by zero. This change does not alter the solutions in the MOM6-examples test cases, but now it permits dimensional rescaling of lengths over a much larger range, and it could change answers if the minimum layer thicknesses are small enough.
Unscale interface heights before taking a global average via a reproducing sum in non-Boussinesq mode global diagnostics to permit dimensional consistency testing over a larger range. All answers are bitwise identical.
Added an optional tmp_scale argument to global_i_mean and global_j_mean to specify an internal rescaling of variables being averaged before the reproducing sum. All answers are bitwise identical, but there are new optional arguments to two public interfaces.
Use tmp_scale when taking the i-mean interface heights for i-mean sponges, to give a greatly expanded range of dimensional consistency testing. All answers are bitwise identical.
MOM6: +(*)Dimensional consistency completion
Refactored how time-averaging of fluxes in forcing types that span multiple timesteps and flux diagnostics are handled, and rescaled the units of fluxes%dt_buoy_accum from [s] to [T]. This involved changing the arguments to fluxes_accumulate, forcing_accumulate, mech_forcing_diags and forcing_diagnostics, but because of the differing types of the arguments, an incompatible mix of code will not compile. Also changed the units of dt as passed to accumulate_net_input, and made a minor change to extractFluxes1d to avoid the possibilty of a division by zero. All answers are bitwise identical, but there are public interface changes, including changes that impact the mct and nuopc driver codes.
Added the new runtime parameters KAPPA_SHEAR_ITER_BUG and KD_TRUNC_KAPPA_SHEAR to permit correction of a dimensionally inconsistent expression in the Newton's method solver code of kappa_shear, and to allow the value of shear mixing that is neglected compared with the background mixing to be set at run-time instead of being hard-coded. By default, all answers are bitwise identical, but there are two new runtime parameters and the MOM_parameter_doc files change.
Added the new runtime parameter VERT_FRICTION_2018_ANSWERS that avoids the use of the hard-coded maximum viscous mixing length per timestep in the vertical viscosity code, and added h_neglect in the denominators of several terms in the viscosity code. All answers in the MOM6-examples test cases are bitwise identical, but the answers will change if ANGSTROM is set to 0, and there is a new entry in the MOM_parameter_doc files.
…AA/MOM6 into Hallberg-NOAA-revise_vert_friction
…OAA/MOM6 into Hallberg-NOAA-simplify_forcing_time
|
I approve this PR, though I do feel left out. It changes answers (ever so slightly) for the dumbbells and the seamounts. |
|
works fine on NCEP WCOSS machine, testing on HEAR now (may need to wait for 1-2 days as HEAR is unstable today) |
|
I approve this PR. Alan.
…On Wed, Dec 18, 2019 at 10:44 PM jiandewang ***@***.***> wrote:
works fine on NCEP WCOSS machine, testing on HEAR now (may need to wait
for 1-2 days as HEAR is unstable today)
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<https://github.com/NOAA-GFDL/MOM6/pull/1046?email_source=notifications&email_token=AHBPTUMWVTTYLS3SQFP5TCLQZLU23A5CNFSM4J34UCBKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEHIJTEI#issuecomment-567318929>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AHBPTUMH4TJFNIW5ATB2ZKDQZLU23ANCNFSM4J34UCBA>
.
|
|
works fine on NCEP HERA machine, answer changes, approve this PR. |
|
Mct tests are failing due to an error in mct_driver/mom_ocean_model_mct.F90, see The equivalent call in coupled_driver does not have this extra argument. We will wait for @Hallberg-NOAA's confirmation before deleting it. |
Corrected arguments to iceberg_forces and iceberg_fluxes in the mct version of update_ocean_model. This should correct the recently introduced problems with compiling MOM6 with the mct coupler.
|
We approve this PR. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This is the PR onto dev/master promised on last weeks dev call. The main changes are:
The parameter documentation is updated so I've pushed a corresponding branch to MOM6-examples: dev-master-candidate-2019-12-17
The length of the list of PRs is misleading. I suspect almost every file has been touched at some point since the last PR to dev/master. There are 365 commits hidden in here:
#1040 from Hallberg-NOAA/rescale_diag_fix
#1039 from adcroft/fix-units-for-meke-advection
#1037 from Hallberg-NOAA/revise_OBCs
#1031 from marshallward/tc4_fix
#1034 from marshallward/adiabatic_clock_id_fix
#1032 from Hallberg-NOAA/rescale_cleanup
#1030 from marshallward/travis_regression_fix
#1019 from Hallberg-NOAA/density_rescale
#1026 from marshallward/nan_merge
#1023 from marshallward/visc_max_chksums
#1021 from marshallward/variable_init_fixes
#1022 from nikizadehgfdl/fix_downsampled_transport_diags
#1025 from marshallward/dev_merge
#1004 from adcroft/tc-reconfig
#1020 from marshallward/mk_fixes
#1018 from marshallward/ale_ineq_1016
#1014 from marshallward/meke_diag_arrays
#1012 from marshallward/omp_test
#1010 from nikizadehgfdl/fix_openmp_compile
#1011 from marshallward/popcnt
#1009 from marshallward/negzero_minmax_log
#1005 from marshallward/make_parallel
#1003 from adcroft/make-label-output
#1002 from marshallward/diag_scaling_p2
#1000 from MJHarrison-GFDL/ALE_sponges_ongrid
#1001 from adcroft/make-src-dependency
#996 from marshallward/codecov_config
#993 from marshallward/travis_stderr
#994 from wfcooke/user/wfc/ePBL_fix
#991 from marshallward/tendency_scaling
#989 from adcroft/tc-reconfig
If you have conflicts it will most likely be due to the dimensional scaling. If you need assistance resolving a conflict reach out to @Hallberg-NOAA .
As usual we'd like reviews from the leads on the major forks: @jiandewang, @awallcraft, @kshedstrom, @gustavo-marques . The latter two have the ability to use the review button - we're trying to get you others that permission too.