Resolve conflicts between ODA use framework #1291 and dev/gfdl #12
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
- Allows us to check that we have not broken the ability to compile with the NUOPC, MCT or coupled_driver modules. - This is not a complete compile but only upto the last module that does not need anything other than MOM6 objects.
This patch reorders many of the loops in horizontal_viscosity in order to improve vectorization of the Laplacian and biharmonic viscosities. Specifically, a single loop containing many different computations were broken up into many loops of individual operations. This patch required introduction of several new 2D arrays. On Gaea's Broadwell CPUs (E5-2697 v4), this is a ~80% speedup on a 32x32x75 `benchmark` configuration. Smaller speedups were observed on AMD processors. On the Gaea nodes, performance appears to be limited by the very large number of variables in the function stack, and the high degree of stack spill. Further loop reordering may cause slowdowns unless the stack usage is reduced. No answers should be changed by this patch, but deserves extra scrutiny given the fundamental role of this function in nearly all simulations.
Added the new module MOM_coms_wrapper, along with explicit interfaces for the broadcast routine for the cases that might actually be used by MOM6. With these new interfaces, the source PE has been made into an optional argument, and there is a new optional argument to indicate whether the broadcast is blocking. Also the MOM_horizontal_regridding module has been updated to reflect these changes. All answers are bitwise identical, but an existing required argument to broadcast has been made optional and there is a new optional argument.
Added the new module MOM_interpolate to wrap the time_interp_external and horiz_interp modules. The overloaded interface time_interp_external had to be renamed to time_interp_extern in the version offered by MOM_interpolate because of a weird compile time problem with the PGI 19.10.0 compiler. Some large blocks of unused code in MOM_horizontal_regridding were commented out. Also modified 6 files outside of the framework directory to use these new interfaces, but they will all work without these changes. All answers are bitwise identical.
- Following suggestion from @marshallward I've implemented an option to select the driver code at the configure stage (changes to configure.ac) - This allows the target build/coupled/ocean_model_MOM.o to be built using the .testing ac tools.
Partially cleaned up MOM_horizontal_regridding.F90, including the elimination of several unused routines, the addition of comments describing variables, and some white space corrections. All answers are bitwise identical.
Moved MOM_diag_manager_wrapper.F90 to MOM_diag_manager.F90 (mostly for brevity) and added use statements and provided interfaces for all of the diag_manager_mod, diag_data_mod and diag_axis_mod routines that are used within the MOM6 code, with some renaming of interfaces to reflect their use within the MOM6 code. All answers are bitwise identical.
Made netcdf module use dependencies explicit in MOM_horizontal_regridding.F90. All answers are bitwise identical.
- Checks that MOM6 can be compiled with the various coupled drivers.
- We spent 1 minute build symmetric unnecessarily when checking the coupled APIs
The fusion of tension and shear strains yields a 1-2% speedup, but also breaks the style convention of capitalization of vertex points, and also evaluates the tension terms over a slightly larger domain, so it has been reverted. A note has been added to investigate this later.
MOM_hor_visc: horizontal_viscosity loop reorder
Nuopc api test
…_more +More MOM6 framework code restructuring
…L/MOM6 into Harrison_ODA_use_framework
marshallward
pushed a commit
that referenced
this pull request
Sep 1, 2021
Merging dev/gfdl 08/05/2021
MJHarrison-GFDL
pushed a commit
that referenced
this pull request
Dec 14, 2021
Bitwise identical refactoring of the code in DOME_initialize_sponges, including renaming variables for greater clarity, adding variables for several dimensional constants, and correcting comments. This also includes more careful handling of the DOME OBCs in DOME_set_OBC_data() to hopefully avoid some obvious problems (noted in a comment about a "fight with T,S") that would arise if a DOME case were set up that used temperature and salinity. Future revisions should add more runtime parameters to specify the details of this case, but properly doing so would involve changing the order of arithmetic; this has not happened in this case. All real variables in this module are now described in comments. All answers and output are bitwise identical. Co-authored-by: Marshall Ward <marshall.ward@noaa.gov>
MJHarrison-GFDL
pushed a commit
that referenced
this pull request
Aug 9, 2022
merge dev/emc into ocean stochastic branch
MJHarrison-GFDL
pushed a commit
that referenced
this pull request
Aug 9, 2022
Note that most of these commits are from previously squashed pull requests, and this PR is restoring them. - 6360dbb Merge branch 'main' into main_to_dev - bac8031 Merge pull request mom-ocean#1566 from jiandewang/EMC-FMS-mixed-mode-20220411 - e532d86 Merge pull request mom-ocean#88 from marshallward/missing_attrib_with_class_bugfix - d380f1d An alternate fix to class(*) issues with FMS 2022-01 - 8ecf333 Merge pull request mom-ocean#87 from jiandewang/feature/update-to-main-20220317 - ba37f94 Merge remote-tracking branch 'FSU/main' into feature/update-to-main-20220317 this is corresponding to MOM6 main 20220317 commit (hash # 399a7db) - 44313d9 Merge pull request mom-ocean#85 from jiandewang/feature/update-to-main-20220217 - 966707f Merge remote-tracking branch 'GFDL/main' into feature/update-to-main-20220217 this is corresponding to MOM6 main branch 20220217 commit (hash # 6f6d4d6), which originally based on GFDL-candidate-20220129 - 32c0e1e Merge pull request mom-ocean#81 from jiandewang/feature/update-to-main-20211220 - 9642b1d delete external/OCEAN_stochastic_phyiscs directory as Phil re-coded in external/stochastic_physics directory - e7c9ada solve minor conflict in mom_cap.F90 mom_ocean_model_nuopc.F90 and MOM_energetic_PBL.F90, add two new files: src/parameterizations/stochastic/MOM_stochastics.F90 and config_src/external/stochastic_physics/stochastic_physics.F90 - 90d5961 Merge pull request mom-ocean#78 from jiandewang/feature/update-to-GFDL-20211019 - fd02017 Merge remote-tracking branch 'GFDL/main' into feature/update-to-GFDL-20211019 - 36f17eb Merge pull request mom-ocean#72 from pjpegion/ocn_stoch_july2021 - a9a957e return a more accurate error message in MOM_stochasics - 56bb41e Merge branch 'ocn_stoch_july2021' of https://github.com/pjpegion/MOM6 into ocn_stoch_july2021 - ca2ae1c update to dev/emc - 14ca4a1 Merge pull request mom-ocean#76 from jiandewang/feature/update-to-GFDL-20210914 - 29016c2 Merge remote-tracking branch 'GFDL/main' into feature/update-to-GFDL-20210914 merge GFDL main 20210914 commit (hash # c09e199) - a8577df Merge branch 'NOAA-EMC:dev/emc' into ocn_stoch_july2021 - f8a8e4c update to gfdl 20210806 (mom-ocean#74) - 16e6af0 update to dev/emc - 237a510 add comments - 1b4273d revert logic wrt increments - 5b2040e add logic to remove incrments from restart if outside IAU window - c5f2b72 add write_stoch_restart_ocn to MOM_stochastics - bdf2dc7 doxygen cleanup - 8bc4acc move stochastics to external directory - a3fa3a1 Merge remote-tracking branch 'upstream/dev/emc' into ocn_stoch_july2021 - e4bc007 stochastic physics re-write - 202cbd4 update to dev/emc - 61717ee Merge remote-tracking branch 'origin/dev/emc' into ocn_stoch - 565e0bb remove debug statements - a4c0411 Merge remote-tracking branch 'upstream/dev/emc' into ocn_stoch - 689a73f remove PE_here from mom_ocean_model_nuopc.F90 - 8afe969 clean up of mom_ocean_model_nuopc.F90 - 25ed4fc revert MOM_domains.F90 - b8d9888 place stochastic array in fluxes container and make SPPT specific arrays allocatable - d984a7e remove stochastics container - eb88219 clean up of code for MOM6 coding standards - 6e3ea1b correct coupled_driver/ocean_model_MOM.F90 and other cleanup - 0b99c1f make stochastics optional - 85023f8 Merge remote-tracking branch 'upstream/dev/emc' into ocn_stoch - 80f9f44 clean up MOM_domains - 5443f8e remove blank link in MOM_diagnostics - 1727d9a re-write of stochastic code to remove CPP directives - 600ebf9 Merge remote-tracking branch 'upstream/dev/emc' into ocn_stoch - 6bb9d0b fix non stochastic ePBL calculation - 1d7ffa3 clean up code - 040e1f1 Merge pull request #13 from NOAA-EMC/dev/emc - 2cba995 Merge branch 'dev/emc' into ocn_stoch - 1dc0f4f Merge remote-tracking branch 'upstream/dev/emc' into dev/emc - 4bd9b9e clean up debug statements - 25ed5ef additions for stochy restarts - a2a374b add stochy_restart writing to mom_cap - 0c15f4c Update MOM_diabatic_driver.F90 - 167a62e Merge pull request #12 from pjpegion/dev/emc - bd477a9 Update MOM_diabatic_driver.F90 - 7212400 Update MOM_diabatic_driver.F90 - 7de295c cleanup of code and enhancement of ePBL perts - cd06356 Merge pull request #11 from NOAA-EMC/dev/emc - 9896d61 Merge pull request #9 from pjpegion/dev/emc_merge - 0a62737 Merge branch 'ocn_stoch' into dev/emc_merge - 3cad1ba Merge pull request #8 from NOAA-EMC/dev/emc - c2aa2a8 updates from dev/emc - 182ef34 additions for stochastic physics and ePBL perts - 671c714 Merge pull request #1 from NOAA-EMC/dev/emc
MJHarrison-GFDL
pushed a commit
that referenced
this pull request
Aug 9, 2022
* initial hooks for stochastic EOS modifications * remove debug statements * add documentation * Change ampltiude from 0.39 to sqrt(.39) * remove global_indexing logic from stoch_eos_init * switch to using MOM_random and add restart capability * update random sequence to update each each time-step * remove tseed0 from MOM_random (leftover from debugging) * Added necessary submodules and S^2, T^2 diagnostics to MOM_diagnostics * Added diagnostics for outputting variables related to the stochastic parameterization. * Diagnostics in MOM_PressureForce_FV updated for stochastic (rather than deterministic) Stanley SGS T variance parameterization. * Added parentheses for reproducibility. * Changed diagnostics to account for possible absence of stoch_eos_pattern in MOM_PressureForce_FV, when deterministic parameterization is on. * remove mom6_da_hooks and geoKdTree from pkg * add stochastic compoment to MOM_thickness_diffuse * fix array size declaration and post_data * Corrected indexing of loops in MOM_calc_varT * Changed how parameterization of SGS T variance (deterministic and stochastic) is switched on in PGF and thickness diffusion codes * Corrected a few typos * Cleaned up indices, redundant diagnostic, printing * Fixed diagnostic IDs * Fixed diagnostics typo * Corrected indices in calculation of tv%varT * Minor index fix * Corrected bug in pressure in Stanley diagnostics * Fixed whitespace error * Stoch eos clock (#5) *Added a clock for the Stanley parameterization Co-authored-by: jkenigson <jkenigso@gmail.com> * add halo update to random pattern * Update MOM_stoch_eos.F90 Fix bug for looping over compute domain (is -> isc etc.) * Avoid unnessary computations on halo (MOM_stoch_eos) and code clean-up (MOM_thickness_diffuse) * Removed halo updates before determ param calc * Update MOM_stoch_eos.F90 Removed unnecessary code * Bug - indices are transposed * Changed Stanley stochastic coefficient from exp(X) to exp(aX) (#9) * Changed Stanley stochastic coefficient from exp(X) to exp(aX) * Extra spaces removed * Stoch eos init fix (#10) * Don't bother calculating tv%varT if stanley_coeff<0 * Missing then added * Merge Ian Grooms Tvar Discretization (#11) * Update MOM_stoch_eos.F90 In progress updating stencil for$ | dx \times \nabla T|^2$ calculation * New discretization of |dx\circ\nablaT|^2 Co-authored-by: Ian Grooms <ian.grooms@colorado.edu> * Multiplied tvar%SGS by grid cell thickness ratio * Added limiter for tv%varT * Stoch eos ncar linear disc (#12) * Update MOM_stoch_eos.F90 In progress updating stencil for$ | dx \times \nabla T|^2$ calculation * New discretization of |dx\circ\nablaT|^2 * AR1 timescale land mask Adds land mask to the computation of the AR1 decorrelation time * Update dt in call to MOM_stoch_eos_run The call to `MOM_stoch_eos_run` (which time steps the noise) is from within `step_MOM_dynamics`. `step_MOM_dynamics` advances on time step `dt` (per line 957), but the noise is updated using `dt_thermo`. It seems more appropriate to update the noise using `dt`, since it gets called from within `step_MOM_dynamics`. * Fixed the units for r_sm_H * Remove vestigial declarations The variables `hl`, `Tl`, `mn_T`, `mn_T2`, and `r_sm_H` are no longer used, so I removed their declarations and an OMP private clause Co-authored-by: Ian Grooms <ian.grooms@colorado.edu> * Update MOM_thickness_diffuse.F90 Changed index for soft convention * Update CVMix-src * Ensure use_varT, etc., initialized * Don't register stanley diagnostics if scheme is off * Stanley density second derivs at h pts (mom-ocean#15) * Change discretization of Stanley correction (drho_dT_dT at h points) * Limit Stanley noise, shrink limiting value * Revert t variance discretization * Reverted variable declarations * Stanley scheme in mixed_layer_restrat, vert_fill in stoch_eos, code cleanup (mom-ocean#19) * Test Stanley EOS param in mixed_layer_restrat * Fix size of TS cov, S var in Stanley calculate_density calls * Test move stanley scheme initialization * Added missing openMP directives * Revert Stanley tvar discretization (mom-ocean#18) * Perform vertical filling in calculation of T variance * Variable declaration syntax error, remove scaling from get_param * Fix call to vert_fill_TS * Code cleanup, whitespace cleanup Co-authored-by: Jessica Kenigson <jessicak@cheyenne1.cheyenne.ucar.edu> * Use Stanley (2020) variance; scheme off at coast * Comment clean-up * Remove factor of 0.5 in Tvar * Don't calculate Stanley diagnostics on halo * Change start indices in stanley_density_1d * Stanley param in MOM_isopycnal_slopes (mom-ocean#22) Stanley param in MOM_isopycnal_slopes and thickness diffuse index fix * Set eady flag to true if use_stored_slopes is true * Cleanup, docs, whitespace * Docs and whitespace * Docs and whitespace * Docs and whitespace * Whitespace cleanup * Whitespace cleanup * Clean up whitespace * Docs cleanup * use_stanley * Update MOM_lateral_mixing_coeffs.F90 * Adds link to another TEOS10 module * Set Stanley off for testing * Line continuation Co-authored-by: Phil Pegion <38869668+pjpegion@users.noreply.github.com> Co-authored-by: Philip Pegion <Philip.Pegion@noaa.gov> Co-authored-by: Jessica Kenigson <jessicak@cheyenne6.cheyenne.ucar.edu> Co-authored-by: Jessica Kenigson <jessicak@cheyenne3.cheyenne.ucar.edu> Co-authored-by: jkenigson <jkenigso@gmail.com> Co-authored-by: jskenigson <jessica.kenigson@colorado.edu> Co-authored-by: Jessica Kenigson <jessicak@cheyenne1.cheyenne.ucar.edu> Co-authored-by: Jessica Kenigson <jessicak@cheyenne5.cheyenne.ucar.edu> Co-authored-by: Philip Pegion <ppegion@Philips-MacBook-Pro.local> Co-authored-by: Jessica Kenigson <jessicak@cheyenne4.cheyenne.ucar.edu>
MJHarrison-GFDL
pushed a commit
that referenced
this pull request
May 20, 2024
Replace db array default values with real literals
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 PR resolves the conflicts between PR#1291 and dev/gfdl. The commits that
are included both here and in PR#1291 are: