Generalized indices: `mpp global field` updates by J-Lentz · Pull Request #1806 · NOAA-GFDL/FMS

J-Lentz · 2025-12-15T18:19:56Z

Description
This PR adds mpp_global_field updates for generalized indices, and causes the unit tests added in #1763 to pass. The MPP_DO_GLOBAL_FIELD_3D_ subroutine is renamed to MPP_GLOBAL_FIELD_, which now takes assumed-rank local and global arguments. Optional xdim and ydim arguments have also been added, which can be used to specify which dimensions are domain-decomposed. MPP_DO_GLOBAL_FIELD_A2A_3D_ is removed, as are the various wrapper subroutines for mpp_global_field (MPP_GLOBAL_FIELD_2D_, MPP_GLOBAL_FIELD_4D_, etc). This PR also adds mpp_domains_mod, which contains new subroutines for packing/unpacking 2D-5D arrays.

How Has This Been Tested?
Builds on C5 with intel-classic. mpp_global_field unit tests pass.

Checklist:

My code follows the style guidelines of this project
I have performed a self-review of my own code
I have commented my code, particularly in hard-to-understand areas
I have made corresponding changes to the documentation
My changes generate no new warnings
Any dependent changes have been merged and published in downstream modules
New check tests, if applicable, are included
make distcheck passes

Refactor the mpp_global_field unit tests to test nonstandard dimension orders. These new tests do not currently pass, as the mpp_global_field implementation uses hardcoded dimension orders.

Test the `mpp_*_group_update` subroutines with nonstandard dimension orders. A separate module for shared unit test code has also been created.

Modify the interpolator tests to test all possible dimension orders.

The 2D tests in test_data_override_ongrid.F90 have been modified to test for generalized indices support.

Update mpp_global_field unit tests to pass `xdim` and `ydim` arguments. Fix issues with the rewritten mpp_global_field that cause build errors and test failures.

Create a new file for `arr2vec`, `vec2arr`, and `arr_init`. Add interfaces for these new subroutines to mpp_domains_mod.

…lobal_field_impl

mlee03 · 2026-02-03T13:25:26Z

+!* governing permissions and limitations under the License.
+!***********************************************************************
+
+      !> Pack an array into a vector


Hi, here to make annoying documentation requests for the chat🤖
Could there be a bit more explanation on why packing is needed and where it's used?

I've added two sentence explanations of packing and unpacking in f42babd.

mlee03 · 2026-02-03T13:29:11Z

+      subroutine ARR2VEC_ (arr, vec, ix, iy, is, ie, js, je)
+        MPP_TYPE_, intent(in) :: arr(..) !< The array to be packed
+        MPP_TYPE_, intent(out) :: vec(:) !< The vector to copy the data into
+        integer, intent(in) :: ix, iy !< Indices of the domain-decomposed dimensions


Just repeating what was mentioned at yesterday's meeting about ix, iy being unclear. Maybe the word axes is clearer? An example would be great

Changed to xdim and ydim in f42babd.

mlee03 · 2026-02-03T13:29:42Z

+        integer, allocatable, dimension(:) :: lb, ub
+        integer :: n, m
+        integer :: i1, i2, i3, i4, i5
+


short explanation of what lb and ub?

Added in f42babd.

mlee03 · 2026-02-03T13:29:49Z

+        ub(ix) = ie
+        ub(iy) = je
+
+        m = 0


comment this block of code

I've added an explanation to the declaration of lb and ub, which I think explains what this block is doing.

mlee03 · 2026-02-03T13:30:10Z

+      end subroutine ARR2VEC_
+
+      !> Unpack a vector into an array
+      !> @ingroup mpp_domains_mod


Similar requests as ARR2VEC_ above

Analogous comments added in f42babd.

mlee03 · 2026-02-03T13:30:51Z

+
+      !> Initialize an assumed-rank array
+      !> @ingroup mpp_domains_mod
+      subroutine ARR_INIT_ (arr, val)


Comment on when to call this subroutine? When it's needed?

Brief explanation added in f42babd.

mlee03 · 2026-02-03T13:33:37Z

+        MPP_TYPE_, intent(in) :: arr(..) !< The array to be packed
+        MPP_TYPE_, intent(out) :: vec(:) !< The vector to copy the data into
+        integer, intent(in) :: ix, iy !< Indices of the domain-decomposed dimensions
+        integer, intent(in) :: is, ie, js, je !< Starting and ending indices of the x and y dimensions


Likewise, examples.. something like array(is:ie, js:je) for ix=1, iy=2
and array(js:je, is:ie) for ix=2, iy=1?

I've added some examples to the documentation for is, ie, js, je.

…lobal_field_impl

Use a temporary variable to store the `type(c_ptr)` when passing the result of `c_loc()` to `c_f_pointer`.

uramirez8707

Minor question, but this looks to me.

uramirez8707 · 2026-04-16T17:24:38Z

Why are the changes in CMakeLists necessary?

They're not necessary. I did some back-and-forth with Gemini trying to determine the cause of a CI build failure and while none of its suggestions ended up being the actual issue, they do seem like genuine improvements, so I kept them in the PR. I can git revert CMakeLists if anyone disagrees.

Jesse Lentz and others added 8 commits September 3, 2025 15:34

Add unit tests for generalized indices

4b9d897

Refactor the mpp_global_field unit tests to test nonstandard dimension orders. These new tests do not currently pass, as the mpp_global_field implementation uses hardcoded dimension orders.

Break up long lines

dbe5f5b

mpp_*_group_update tests with generalized indices

2b48872

Test the `mpp_*_group_update` subroutines with nonstandard dimension orders. A separate module for shared unit test code has also been created.

interpolator generalized indices tests

2bf6220

Modify the interpolator tests to test all possible dimension orders.

2D generalized indices tests for data_override

7819072

The 2D tests in test_data_override_ongrid.F90 have been modified to test for generalized indices support.

Tentative mpp_global_field updates

39dbf83

Update mpp_global_field unit tests

2e52375

Update mpp_global_field unit tests to pass `xdim` and `ydim` arguments. Fix issues with the rewritten mpp_global_field that cause build errors and test failures.

Move packing/unpacking routines into new file

572f48c

Create a new file for `arr2vec`, `vec2arr`, and `arr_init`. Add interfaces for these new subroutines to mpp_domains_mod.

J-Lentz requested review from bensonr, rem1776, uramirez8707 and vithikashah001 as code owners December 15, 2025 18:19

J-Lentz requested a review from abrooks1085 December 15, 2025 18:20

Jesse Lentz added 2 commits January 20, 2026 14:31

Merge remote-tracking branch 'origin/main' into general_indices_mpp_g…

2ddef53

…lobal_field_impl

Split long line

86abf41

mlee03 suggested changes Feb 3, 2026

View reviewed changes

Jesse Lentz added 8 commits April 2, 2026 12:44

Merge remote-tracking branch 'origin/main' into general_indices_mpp_g…

348d123

…lobal_field_impl

Workaround for GFortran bug

8132eea

Use a temporary variable to store the `type(c_ptr)` when passing the result of `c_loc()` to `c_f_pointer`.

Improve documentation

f42babd

Add target attribute to mpp_domains_stack in non-MPI build

549d1fd

Attempt to fix CMake build

ddea957

Fix CMake build: Attempt 2

5a8f60f

Fix CMake build: Attempt 3

9ab20fc

Use permute_arr in mpp_global_field tests

c95f694

uramirez8707 approved these changes Apr 16, 2026

View reviewed changes

Conversation

J-Lentz commented Dec 15, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

J-Lentz Apr 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

uramirez8707 left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

J-Lentz commented Dec 15, 2025 •

edited

Loading

J-Lentz Apr 9, 2026 •

edited

Loading