Merge MPAS-Dev::develop into release-stable #8
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
…olation
This commit adds three new fields to the Registry.xml file,
bdyMask{Cell,Edge,Vertex}, plus new code to ensure that pixels of static
data that are mapped to a cell actually belong to that cell. This latter change
is needed for limited-area meshes, since the remapping of source pixels for
terrain, land use, soil category, monthly albedo, monthly vegetation fraction,
and maximum snow albedo using a path-based search does not guarantee that
the pixel actually lies in the cell where the search terminates (in particular,
in the case of pixels outside of a limited-area mesh, where the search
terminates at the boundary cell closest to, but not containing, the pixel).
If the bdyMask{Cell,Edge,Vertex} fields are not present in the input file, they
will receive a default value of zero, which will effectively deactivate the new
logic added in this commit, giving results as in prior versions of the code.
At present, a bdyMaskCell value of 7 (via the parameter nBdyLayers) is assumed
to represent the value for cells on the outermost edge of a limited-area mesh;
however, if regional meshes make use of different numbers of relaxation cell
layers or specified cell layers, this will need to be generalized.
Eventually, some of the new functions in this commit should be considered
for migration to mpas_geometry_utils.F.
When applying terrain smoothing on limited-area meshes, we now set the terrain for non-existent neighbor cells to the same value as the terrain in the cell where the terrain filter is being applied.
Similarly to what was done for terrain in the previous commit, update the smoothing of coordinate surfaces so that references to non-existent neighbors yield the same value of hx as in the cell itself. This commit also removes some commented-out, unused logic in the coordinate surface smoothing code.
This commit adds logic to set the index of non-existent cells -- those on the exterior side of a boundary edge in a limited-area mesh -- to the same as the cell that does exist. This is done only to avoid bad values in the computed z-metric terms zb and zb3 for boundary edges. The exact value assumed in the computation of zb and zb3 for edges on the outermost boundary of a limited-area mesh should not matter, since these edges are specified-zone edges that border specified-zone cells.
…Dev#204) This merge enables the production of static fields and vertical grids for limited-area meshes in the init_atmosphere core. * init_atmosphere/regional_static: Avoid unpredictable values in metric terms zb, zb3 for boundary edges Modify coordinate surface smoothing to work with limited-area meshes Modify terrain smoothing to work with limited-area meshes Improvements to init_atmosphere core to enable regional static interpolation
This commit creates an `ISO_C_BINDING` interface for `read_geogrid.c` file. As well, it contains a number of other changes to `read_geogrid.c` and its corresponding callers: `mpas_init_atm_gwd.F` and `mpas_init_atm_static.F`. Per the ISO_C_BINDING interface, the arguments of `read_geogrid.c` were changed from pointers, to variables to remove the need to for dereferencing. As well, the `int len` argument was removed. Instead of prepending a null character at the end of `fname` with C, one can create a null-terminated c string with `mpas_c_to_f_string` routine. In `mpas_init_atm_gwd.F` and `mpas_init_atm_static.F`, a number of changes have occurred to rarray. First, rarray has been set to `pointer` and `contiguous`. Instead of passing the Fortran array by reference to C, we now pass the C address via: `rarray_ptr = c_loc(rarray)`, where `rarray_ptr` is of `type (c_ptr)` in Fortran. Also contains fix (from PR review) to the read_geogrid interface variable fname.
This commit creates an `ISO_C_BINDING` interface for the `pool_hash.c` file. To handle the interface between Fortran and C, a new function (`pool_hash(...)` was created in `mpas_pool_routines.F` to act as a wrapper for the `pool_hash.c` function (`c_pool_hash(...)). The wrapper function converts the Fortran types into corresponding C types and calls `c_pool_hash`. The `len` argument for `c_pool_hash` has been removed and has been replaced with `strlen(key)` accordingly. Note that this works now because of the use of `mpas_f_to_c_string` function module, which prepends a null character onto `key`. As well note, that if it is desired to create a hash that is greater then `strlen(key)`, the length will need to be re-added in another commit. Also contains the fix (from the PR review) to loop until the null character within `key` instead of using `strlen`.
This commit creates two new `ISO_C_BINDING` interfaces for the functions found in `random_id.c`: `seed_random()` and `gen_random()`.
This commit adds an ISO_C_BINDING, 2003 standard compilent interface to two C functions: `compute_ev_2` and `compute_ev_3` within mpas_ocn_okubo_weiss.F. As well, it removes the `ifdef UNDERSCORE` in mpas_ocn_okubo_weiss_egienvalues.c.
This commit removes references to -DUNDERSCORE in the top level Makefile as well as the ACME and CESM Makefile.
This commit adds a new initialization case, case 9, to produce lateral boundary conditions data for regional MPAS-Atmosphere simulations. At present, the new routine, init_atm_case_lbc, is just a placeholder that is called from the main init_atm_setup_case routine when config_init_case == 9 in the namelist file. The init_atm_case_lbc routine is called in a time loop from config_start_time to config_stop_time every config_fg_interval seconds. This commit also updates the description of the config_init_case namelist option in the init_atmosphere core's Registry.xml file to document the existence of init case 9.
This commit defines an 'lbc' output stream for writing model LBC files, an 'lbcs' package to control the writing of the 'lbc' stream and the allocation of the 'lbc_state' variables, and a new var_struct defining the fields that are needed in MPAS-A LBC files. Code has been added to properly set up packages when init case 9 is run, and appropriate calls to the MPAS stream mananger for the 'lbc' stream have also been added.
The init_atm_case_lbc routine now produces valid lbc_u, lbc_w, lbc_theta,
lbc_rho, and lbc_{qc,qv,qr} fields. The implementation of the routine borrows
from the init_atm_case_gfs routine, which produces similar fields for
the model initial conditions. In future, it may be possible to factor out
duplicated code between init_atm_case_lbc and init_atm_case_gfs.
If the config_fg_interval value specified in the namelist.init_atmosphere file doesn't match the output_interval specified for the 'lbc' stream when producing LBCs, stop the init_atmosphere core, as there's generally no good reason for not having these intervals match.
This merge creates ISO_C_BINDING interfaces for C functions in the atmosphere and
ocean core to eliminate the need for the -DUNDERSCORE name mangling logic.
It also updates each C source file to be more standards compliant, for example, by
passing in arrays using the c_loc ISO_C_BINDING function (see, e.g.,
read_geogrid.c). Where necessary, the mpas_f_to_c_string routine from
mpas_c_interfacing is used to create null-terminated strings for C, which
eliminates the need to pass in the length of a string in some circumstances
(specifically, read_geogrid.c and pool_hash.c).
Lastly, many arguments to C routines are passed by value to avoid the need to
deference them in C.
*** Conflicts: There were quite a few conflicts in mpas_init_atm_static.F, and
resolving them involved removing the old code for interpolating VARSSO and
pre-computed GWDO fields, and adding c_loc and mpas_f_to_c_string calls for
MODIS-based max snow albedo, monthly albedo, and monthly green fraction.
* framework/dunderscore:
Remove references to -DUNDERSCORE within Makefiles
ISO_C_BINDING interface for computer_ev_2/3 for Ocean
ISO_C_BINDING Interface for random_id.c
ISO_C_BINDING update for pool hash
ISO_C_BINDING Interface for read_geogrid.c
Conflicts:
src/Makefile.in.CESM
src/core_init_atmosphere/mpas_init_atm_static.F
There were about a dozen write statements in the MPAS-A dynamics that were used for debugging, but can now be safely removed.
This commit removes print statements guarded by an if-test on "debug_regional", and it also removes the debugging routine atm_bdy_check_values.
…imulations The new namelist option, config_apply_lbcs, defaults to .false. and resides in the new &limited_area namelist group. When set to .true., this option activates the code that was previously activated by the 'regional_mpas' parameter at the top of the atm_time_integration module.
…truct The 'limited_area' package is active whenever the config_apply_lbcs namelist option is true. This package controls the reading of the 'lbc_in' stream as well as the allocation of all variables in the 'lbc' var_struct.
…hysics suite The scalars are reset along the domain boundaries individually (i.e., for qv, qc, etc.) after scalar transport and after microphysics. Previously, there was no logic to detect which scalars where actually active and which were not, so it was necessary to comment or un-comment, e.g., the handling of nr and ni, depending on which physics suite was being run. Now, there is a check to see that the index of each scalar is valid before setting that scalar along the domain boundaries. Also, it appears that the last argument to mpas_atm_get_bdy_state in the previously commented-out calls to mpas_atm_get_bdy_state for nr and ni had an incorrect final argument in the call to mpas_atm_get_bdy_state: this argument should be 'dt' rather than 'rk_timestep(rk_step)'.
The 'xtime' field was previously being read by the 'lbc_in' stream, but this is not necessary and in some cases may actually be problematic. While the 'xtime' field must be present in the LBC files to allow the stream manager to seek to the correct LBC update file at any given point in the model simulation, we don't actually need to read the 'xtime' field. In cases where the time in the LBC file being read does not exactly match the current simulation time -- because we don't request an exact time match in the stream manager read calls in mpas_atm_update_bdy_tend -- the xtime field that was read in would overwrite the model xtime field with an incorrect time. The 'xtime' field is set to the correct simulation time in the atm_timestep routine in each timestep, but this happens only after most of the physics (all non-microphysics schemes, currently) are called.
When computing rho_zz, we use array syntax to divide rho(:,:) by zz(:,:). The value of zz in the "garbage cell" is generally not predictable, and this can lead to a divide by zero FPE. To avoid this potential FPE, we now set zz(:,nCells+1) = 1.0 in mpas_atm_update_bdy_tend. In principle this should be completely safe, since no other code should rely on the garbage element for a field taking on any particular value.
…constituents When the mpas_atm_get_bdy_state routine is called for scalar constituents (e.g., qv), the calls to mpas_pool_get_array will not find the LBC field in the tend and state pools as a standard field, leading to warning messages when the code is compiled in debug mode. These messages are expected and harmless, so this commit adds calls to set the error mode for the pool routines to MPAS_POOL_SILENT only for these calls to mpas_pool_get_array.
…efile The module_sf_noah_seaice.o file depends on module_sf_noahlsm.mod. Occasionally, parallel builds of MPAS-Atmosphere would fail due to this missing dependency in src/core_atmoshere/physics/physics_wrf/Makefile. This commit adds a dependency on module_sf_noahlsm.o for module_sf_noah_seaice.o.
…n boundary This commit adds a new function, blend_bdy_terrain, that combines the first- guess terrain field from an intermediate file with the terrain field produced by the init_atmosphere core's "static interpolation" step along the boundary cells of regional domains. The new function is called only when a new namelist option, config_blend_bdy_terrain in the &vertical_grid namelist group, is true. Terrain blending happens before terrain smoothing and vertical grid generation, but after static interpolation. The first-guess terrain is taken to be the SOILGHT field found in an intermediate file described by config_met_prefix and config_start_time. This commit also removes the deactivated code in the real-data test case setup for using the first-guess terrain everwhere.
…ified_physics (PR MPAS-Dev#223) This merge adds a missing dependency for module_sf_noah_seaice.o in physics_wrf/Makefile. The module_sf_noah_seaice.o file depends on module_sf_noahlsm.mod. Occasionally, parallel builds of MPAS-Atmosphere would fail due to this missing dependency in src/core_atmoshere/physics/physics_wrf/Makefile. This commit adds a dependency on module_sf_noahlsm.o for module_sf_noah_seaice.o. * atmosphere/noah_makefile_dependency: Add missing dependency for module_sf_noah_seaice.o in physics_wrf/Makefile
This merge fixes a bug in the PIO compatibility check in the top-level Makefile, wherein the PGI compilers failed to build any of the test programs due to the order in which arguments were provided in the compilation command. The simple fix is to move the source file to the front of the compiler arguments. * framework/pio_pgi_fix: PIO Version Compatibility - PGI Bug Fix
…the logical config_o3climatology from false to true. This change forces the RRTMG long- and short-wave radiation codes to use the monthly-mean global climatological ozone data also used in the CAM radiation code, instead of a single vertical profile of ozone, one for winter and one for summer. When config_o3climatology is true, the three-dimensional ozone data is updated every day using temporal interpolation of the input global climatological data on fixed presssure levels. The three-dimensional data is interpolated to the time-varying pressure levels before calls to the RRTMG radiation codes.
Several loops to enforce LBCs over boundary edges needed to cover owned and halo edges to ensure correct results with multiple MPI tasks, and one OpenMP barrier in the code to adjust scalars along boundaries needed to be activated to ensure correct results with more than one OpenMP thread. The dycore now gives bit-identical results for limited-area domains any MPI task count and any OpenMP thread count, provided compiler flags like "-fp-model precise" are added.
…v6.3 (PR MPAS-Dev#183) This merge sets the use of CAM climatological O3 by default, rather than using a single profile. This change forces the RRTMG long- and short-wave radiation codes to use the monthly-mean global climatological ozone data also used in the CAM radiation code, instead of a single vertical profile of ozone, one for winter and one for summer. When config_o3climatology is true, the three-dimensional ozone data is updated every day using temporal interpolation of the input global climatological data on fixed presssure levels. The three-dimensional data is interpolated to the time-varying pressure levels before calls to the RRTMG radiation codes. * atmosphere/unified_physics.o3_climatology: In ./src/core_atmosphere/Registry.xml, switch the default value of logical ...
added 6 commits
December 10, 2020 11:00
jjguerrette
requested review from
liujake,
byoung-joo,
snyderdad,
svahl991 and
junmeiban
snyderdad
approved these changes
Dec 14, 2020
There was a problem hiding this comment.
I've not looked through the many individual commits here, but I think we have to update to what they have in develop branch of MPAS-Dev.
@jjguerrette can you confirm that with this PR MPAS-Model will now be up to date with develop branch of MPAS-Dev?
Yes, that is what is achieved in commit 5beb733 and after. If you look at the list of commits in the PR, those commits are almost at the bottom of the list. Every commit before that one are the changes made to MPAS-Dev::develop since we performed a similar merge. I'm not sure if we should do a "squash and merge" or a regular merge. A regular merge would allow for the MPAS-Dev history to be combined with ours, but it is not our usual practice. |
My opinion is we should do whatever makes our life easiest. In most other repos, that is "squash and merge". That may not be the case with this repo. |
svahl991
approved these changes
Dec 14, 2020
|
Should I run a short-range cycling experiment to make sure we get similar results? |
liujake
approved these changes
Dec 14, 2020
It would be interesting to know if the "compare" ctests were passing with exact matches or just falling within the accepted tolerances. I'm not sure if there's a simple way to tell that. If the results are matching exactly for gnu-openmpi, then I think a cycling test is probably not needed. It would normally be fairly simple for me to force a run of the automated cycling test and we could compare cycling results that way. But right now that test isn't working. (JCSDA-internal/mpas-jedi#464) |
|
That makes sense Steve. I can build and compare with a zero tolerance for the non-forecast applications. If I go to try to do a cycling test, I may run into the same error message that you see in the the automated test. Conversely, it might help your investigation of the error if I try it. |
|
Yes, please. |
|
I pinged Michael Duda about other pending mods listed in this issue. Plan is to start an issue in MPAS-Dev containing that list, then link to PRs as they made and check off items as they are merged. |
|
Full month cycling yields numerically identical results. See here: https://github.com/JCSDA-internal/mpas-jedi/pull/456#issuecomment-757007159. Ready to merge pending remaining clang tolerance failure for AD in linear getvalues test. |
Refers to https://github.com/JCSDA-internal/mpas-jedi/pull/192? |
I think it is this commit actually: b22c340 PR was in my fork here: jjguerrette#10 Here is the feature branch: https://github.com/jjguerrette/MPAS-Model/tree/feature/add_sfvp |
8 tasks
This was referenced Feb 17, 2021
Closed
climbfuji
pushed a commit
that referenced
this pull request
Aug 26, 2022
Merge latest changes from upstream repository. Works with mpas-jedi branch of the same name. See that corresponding PR for more information on mpas-bundle behavior. There are a lot of commits in this PR due to it having been a long time since such a merge was conducted. The conflicts between the two branches are dealt with in commits 5beb733 and after. Some small code changes from the early days of MPAS-JEDI development is removed, because it no longer serves a purpose. After this PR is merged, `release-stable` will be nearly up to date with MPAS-Dev::develop with the following exceptions for code that is only in the JCSDA-internal fork: - convective/cloud diagnostics for PANDA-C - 2-stream I/O with a static file shared across all dates - adjoint of some halo communications used for static B - bug fixes in core_init_atmosphere for initializing `rho_zz` and `xice` - optional cmake build process and dependence on the `jedi-cmake` repository Closes #7
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.
Merge latest changes from upstream repository. Works with mpas-jedi branch of the same name. See that corresponding PR for more information on mpas-bundle behavior.
There are a lot of commits in this PR due to it having been a long time since such a merge was conducted. The conflicts between the two branches are dealt with in commits 5beb733 and after. Some small code changes from the early days of MPAS-JEDI development is removed, because it no longer serves a purpose.
After this PR is merged,
release-stablewill be nearly up to date with MPAS-Dev::develop with the following exceptions for code that is only in the JCSDA-internal fork:rho_zzandxicejedi-cmakerepositoryCloses #7