CABLE Library for AM3/ACCESS3 by Whyborn · Pull Request #724 · CABLE-LSM/CABLE

Whyborn · 2026-05-22T04:33:36Z

CABLE

Thank you for submitting a pull request to the CABLE Project.

Description

This builds the library for AM3 and future ACCESS3 applications. It requires a rejig of the original library, taking some files out of the default library and including them only in ESM1.6. The intention was that the library would only be the science code, but this is not possible with the current setup.

A summary of things that the science directory depends on, that extends outside the science directory that I've identified:

The biogeophysics derived types are not defined within the science directory, as opposed to the CASA derived types which are defined within the science directory. Each of these have different definitions for each application (offline, ESM1.6, AM3). Logically, there should only be one definition, but I don't think this will happen any time soon. In the mean time, including the type modules in src/coupled/AM3/control/cable/CM3 in the library should be an acceptable solution.
The trickier problems are the namelists and surface types handling. These are related issues, as the cable_surface_types_mod module also reads the namelist that sets the CABLE surface types. Do the surface types need to be handled in this way? Is there any foreseeable situation where anyone would not use the standard CABLE PFT indexing? If we remove this, it makes things significantly easier. It also does the extra work of defining a bunch of required JULES variables. I think the best solution is go back to hard-coded CABLE PFT indices, see this JULES PR.

Testing

Are the changes bitwise-compatible with the main branch? If working on an optional feature, are the results bitwise-compatible when this feature is off? If yes, copy benchcab output showing successful completion of the bitwise compatibility tests or equivalent results below this line.

📚 Documentation preview 📚: https://cable--724.org.readthedocs.build/en/724/

Whyborn · 2026-07-03T06:48:02Z

This now, along with JULES PR #346, access-spack-packages PR #455 and UM PR #148 successfully builds AM3 with CABLE as a library and achieves bitwise compatible results.

ccarouge · 2026-07-06T06:16:04Z

@Whyborn Can you please add the benchcab results for bitwise reproducibility?

ccarouge

A few requests for more explanation and some relatively minor changes.

ccarouge · 2026-07-06T06:19:14Z

      spack-manifest-data-path: .github/build-ci/data/standard.json
      # We need to install the compilers manually here because the upstream image is too large to run on GitHub-hosted runners
      spack-compiler-manifest-path: .github/build-ci/compilers/spack.yaml
+      access-spack-packages-ref: 1d9acf45c39229ff4e90a9aa49922b1169305476


What is this line doing? Why is it needed now and not before?

That's to make the CI use the correct version of the spack packages. I should get the spack packages PR merged in first, tag it, then refer to that tag here.

ccarouge · 2026-07-06T06:20:21Z

Why is this file modified for the library work? I would guess we do not compile this for AM3.

This was intended to be a bit of future proofing, as Sean's MPI slightly modified the land use change code which was referred to by the AM3 code, even if not yet active. We came up with a better solution in this merged PR. I'll remove this change.

ccarouge · 2026-07-06T06:22:30Z

Why are we adding this to science/ when it is only to be used in ESM1.5/1.6? Should the # define ESM15 YES be # define ESM15 NO in the offline repository? Or even, remove that line from the file and deal with it at compilation as usually done for pre-processor flags.

And why is it part of the AM3 library work?

Unfortunately the land use code had moved in a bit from the form that is in AM3. The JULES src/science_cable/landuse/landuse3.F90 should be the same as src/science/landuse/landuse3, but it isn't, and the coupling code is written based on the JULES version. A few routine signatures have changed, and I was getting type mismatch errors when I tried to quickly fix it.

I should remove the #ifdef and #def. I think it's best to get this in first, then reconcile the differences between the way this routine is called in the coupling and then remove this file.

ccarouge · 2026-07-06T06:25:08Z

+	src/science/landuse/landuse3.F90
+        src/science/landuse/landuse_constant.F90
        src/params/cable_phys_constants_mod.F90


Could we put this back in alphabetical order? And with the correct indentation.

Yep, will fix that.

ccarouge · 2026-07-06T06:26:40Z

-        src/science/casa-cnp/casa_dimension.F90
        src/science/casa-cnp/casa_feedback.F90
        src/science/casa-cnp/casa_inout.F90
        src/science/casa-cnp/casa_param.F90
-        src/science/casa-cnp/casa_phenology.F90
-        src/science/casa-cnp/casa_readbiome.F90
        src/science/casa-cnp/casa_rplant.F90
        src/science/casa-cnp/casa_sumcflux.F90
-        src/science/casa-cnp/casa_variable.F90


What happens to all these CASA files for the ESM1.6 library? Were they added but are in fact not needed?

Unfortunately these are bits that had been rewritten for AM3- the way the derived types are structured is different to offline, see this JULES issue. This is related to the duplicate ALLOCATABLE/POINTER type definitions that I wanted to unify some time back. The AM3 versions of these live in the JULES src/util/casa.

ccarouge · 2026-07-06T06:29:56Z

+        if (CMAKE_Fortran_COMPILER_ID MATCHES "Intel")
+            add_compile_options(-r8 -i8)
+        elseif(CMAKE_Fortran_COMPILER_ID MATCHES "GNU")
+            add_compile_options(-fdefault-real-8 -fdefault-integer-8 -fdefault-double-8)
+        else()
+            message(STATUS "Cannot compile library with unknown compiler")
+        endif()


These are the same options for ESM1.6 and ACCESS3. Are we doubling them in case we want to modify them for ACCESS3? I would prefer to put this block under the if(CABLE_LIBRARY) condition directly and have them appear only once. So that it's clearer the only difference between ESM1.6 and ACCESS3 libraries are the sources.

I think it makes more sense to declare them separately- it makes it clearer, and there may be instances where the flags are not the same and this preempts that.

Whyborn · 2026-07-09T06:05:08Z

Benchcab output:

2026-07-09 15:54:21,176 - INFO - benchcab.benchcab.py:391 - 0 failed, 168 passed

Whyborn and others added 30 commits May 22, 2026 14:19

Rejig library config for AM3

785febe

Add library manifests to CI

5e0bd5a

Add library manifests to CI

5ea70cd

Set ref for spack-packages

1bdd923

Remove templates taken from spack-packages

3cb4057

Fix quotes

329c302

Misspelled matches

4a692ca

Move some files to common library

395b1fa

Add type files to library

d033810

Fix some file names

f84c396

Fix more file names

fea2e45

Drop coupled namelists mod

b5e9b4d

Add surface types mod

367ba2f

Fix name

878632d

use new cable_surface_types_mod.F90

114cf54

Wrong end subroutine name

cb6ac5c

Add stash mapping

26c5d59

Add ncpft back

8b94bfb

Add dimension declaration

6f9b9b0

Remove CASA IO from library

0bd7975

Readd landuse

de283d4

Fix file name

921abd5

Add cable_iovars.F90 to the build with slight adjustment

feb632b

Readd casa_inout

dc432f7

Library now builds

4813a19

Remove casa_variable.F90 from library

911c2b3

Add all the CASA types

f5662be

Update CASA types

c6fe3e5

Use new cable_surface_types_mod.F90

8d4ebc6

Fix dummy argument intents

6d8f1a1

Whyborn and others added 17 commits June 24, 2026 11:37

Readd casa_inout

2df4cd6

Library now builds

7c485cc

Remove casa_variable.F90 from library

5fe937f

Add all the CASA types

3d5343c

Update CASA types

084d900

Use new cable_surface_types_mod.F90

f845f2e

Fix dummy argument intents

d8803a2

Accidental doubly defined argument

cd5f58f

Fix typos

0672feb

Remove IFDEF statement

95c8a00

Remove duplicate files

b9ba147

Set ncpft

3e2f3cf

Remove coupled version of cable_surface_types

4f0d020

Fix name error

174ab43

Update cable_user and CMakeLists.txt

79fdbd3

Include ESM1.5 version of casa_landuse

595bce0

Add missing file

3ef0fed

Whyborn force-pushed the make-AM3-library branch from aaba5ca to 3ef0fed Compare June 24, 2026 01:38

Whyborn added 2 commits July 3, 2026 12:55

Update science

2f78493

ff pull

43bd6c6

Whyborn mentioned this pull request Jul 3, 2026

Adjustments to UM and CABLE spack packages for CABLE library ACCESS-NRI/access-spack-packages#455

Open

Whyborn requested a review from ccarouge July 3, 2026 07:05

Whyborn added 2 commits July 3, 2026 17:06

Merge branch 'main' into make-AM3-library

e413350

Didn't fix the merge conflict properly

7cd8286

ccarouge requested changes Jul 6, 2026

View reviewed changes

Address PR feedback

ef8cdfd

Whyborn requested a review from ccarouge July 9, 2026 06:25

Uh oh!

Conversation

Whyborn commented May 22, 2026 • edited by github-actions Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

CABLE

Description

Testing

Uh oh!

Whyborn commented Jul 3, 2026

Uh oh!

ccarouge commented Jul 6, 2026

Uh oh!

ccarouge 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!

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!

Whyborn commented Jul 9, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Whyborn commented May 22, 2026 •

edited by github-actions Bot

Loading