#1312 ScalarArray code generation #3101

mo-alistairp · 2025-08-19T10:30:39Z

Adding code generation for ScalarArrays (towards #1312 ). The metadata support was added in #2173.

arporter · 2025-08-19T12:24:31Z

src/psyclone/domain/lfric/kern_call_acc_arg_list.py

+    def scalar_array(self, scalar_arr_arg, var_accesses=None):
+        '''
+        Override the default implementation as there's no need to specify
+        ScalarArrays for an OpenACC data region.


This isn't true I'm afraid - they will need to be added to a data region. However, you could leave this as a TODO as I'm not sure we're ever going to need data regions.

Thanks for that - it's currently a very rough draft so some of the comments are currently just borrowed from other functions. But, I will definitely note that down!

codecov · 2025-08-19T16:11:25Z

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.95%. Comparing base (e9d4423) to head (f187466).

❌ Your project check has failed because you have indirect coverage changes. Learn more about Unexpected Coverage Changes and reasons for indirect coverage changes.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #3101      +/-   ##
==========================================
- Coverage   99.95%   99.95%   -0.01%     
==========================================
  Files         378      379       +1     
  Lines       53723    53822      +99     
==========================================
+ Hits        53701    53799      +98     
- Misses         22       23       +1

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

…erly

mo-alistairp · 2025-10-21T15:50:33Z

This still isn't working as intended, but is getting closer to correct. As an example, the code gen is now adding the scalar arrays to the generated code, alongside the arrays containing their dimensions. However, it is not adding these dimensions to the generated code (e.g., no dimensions(...) mention). It is also assigning the field as intent(in) right now, which it shouldn't do either.

Furthermore, the stub_gen test is failing because we are currently adding the dimensions array to the psyir layer as an ArgumentInteraface which isn't there as an argument at the call. Need to find a way around this. I'm wondering if we need a dimensions array at all but I'll have to investigate that further.

mo-alistairp · 2025-12-19T09:54:57Z

I'm having a little trouble getting testkern_scalar_array_mod via lfric_scalar_codegen_test to compile. Currently, it's erroring with
Error: Symbol 'gh_scalar_array' at (1) has no IMPLICIT type; did you mean 'gh_scalar'?

Still trying to find what needs adding to fix this.

mo-alistairp · 2026-01-16T12:13:32Z

This is now working and passing the tests with and without compilation. ~~I'm not sure why my commits have started coming up as unverified, but I'll try and fix that.~~

I'll warn you that I've had to do a slightly strange fix to correct the ordering of the declarations in code generation. As mentioned on Monday, there isn't a way of changing the order of declarations - they appear in the order that they are added to the symbol table. Since the ScalarArray is an argument it was always being added before the dimensions array, and this was causing compile time errors as the dimensions array is being used in the ScalarArray declaration. My fix for this is simply finding the ScalarArray in the symbol table and replacing it with itself, so that it is being added after the dimensions array. Hopefully you're happy with this

This should now be ready for review again

mo-alistairp · 2026-01-16T16:23:40Z

I'm not sure why CodeCov has only just decided that this was a problem. I've looked at the line it's complaining about (line 773 in arg_ordering.py) and I don't see an easy way of avoiding this problem without splitting scalar_args out of the scalar function. The test for this is checking whether the argument isn't is_scalar or is_scalar_array, which I can't think of a way of doing independently without splitting the function into two. I don't think we want to do this from a design perspective.

I still think that this is ready for review, and would be interested if you have any thoughts about this

arporter

Looking good @mo-alistairp - well done!
There's a bit of tidying-up still to do and I'm asking that you make the test case a bit more involved. I suspect what you have is vulnerable to a name clash so you'll see I've asked you to switch to using a tag for the new variables.
I'll fire off the integration tests in the meantime.

arporter · 2026-01-19T08:23:02Z

src/psyclone/domain/lfric/kern_call_acc_arg_list.py

            :py:class:`psyclone.core.VariablesAccessMap`]

        '''
+        # TODO: Add implementation of OpenACC data region for ScalarArrays


Please remove the TODO as otherwise our rule is that you need an associated Issue and I don't think we want another Issue :-)

arporter · 2026-01-19T08:23:36Z

src/psyclone/domain/lfric/kern_call_acc_arg_list.py

            :py:class:`psyclone.core.VariablesAccessMap`]

        '''
+        # TODO: Add implementation of OpenACC data region for ScalarArrays


Also, please document the NotImplementedError as a ":raises ...." in the docstring.

arporter · 2026-01-19T08:29:15Z

src/psyclone/domain/lfric/lfric_scalar_array_args.py

    Handles the declarations of ScalarArray kernel arguments appearing in
    either an Invoke or a Kernel stub.

-    :param node: the Invoke or Kernel stub for which to manage the \


Please put back the description of the node argument.

arporter · 2026-01-19T09:45:28Z

src/psyclone/domain/lfric/lfric_scalar_array_args.py

+        super().__init__(node)
+
+        # Initialise dictionaries of 'real', 'integer' and 'logical'
+        # ScalarArray arguments by data type and intent


Thanks, please update the comment.

arporter · 2026-01-19T10:18:10Z

src/psyclone/domain/lfric/lfric_scalar_array_args.py

-            if self._logical_scalar_arrays[intent]:
-                for arg in self._logical_scalar_arrays[intent]:
+            for arg in self._scalar_array_args[intent]:
+                if arg.intent in FORTRAN_INTENT_NAMES:


Do you need this if? If so, please add a comment.

arporter · 2026-01-19T12:18:56Z

src/psyclone/tests/test_files/lfric/28.scalar_array_invoke.f90

+  integer(kind=i_def)                         :: a_scalar
+
+  call invoke(                                                                            &
+       testkern_scalar_array_type(afield,real_array,logical_array,integer_array,a_scalar) &


Probably worth making this more complicated by including a second kernel call (to one of the existing test kernels) in the invoke. If you could make one of the variable names such that you would get a clash with the dims_real_field name that PSyclone generates then that would be great. (e.g. you could re-name "a_scalar" to "dims_real_field")

arporter · 2026-01-19T13:52:41Z

src/psyclone/domain/lfric/lfric_scalar_array_args.py

+            if scal_multi_type:
+                raise GenerationError(
+                    f"ScalarArray argument(s) {scal_multi_type} in Invoke "
+                    f"'{self._invoke.name}' have different metadata for data "


perhaps replace everything after and including "have" with "is/are passed to more than one kernel and the kernel metadata for the corresponding arguments specifies different intrinsic types."? (I really struggled to understand this message at first.)

arporter · 2026-01-19T13:56:01Z

src/psyclone/domain/lfric/lfric_scalar_array_args.py

+        for intent in FORTRAN_INTENT_NAMES:
+            for arg in self._scalar_array_args[intent]:
+                # Distinguish whether they are ScalarArrays
+                if (arg.descriptor.data_type not in


I think it would be simpler to put this check at L150 - i.e. before we add arg into the _scalar_array_arg dictionary. That would save us having to loop through again to check what we've just added.

arporter · 2026-01-19T13:57:04Z

src/psyclone/domain/lfric/lfric_scalar_array_args.py

+                        f"{const.VALID_SCALAR_DATA_TYPES}.")
+
+        # Create declarations
+        self._create_declarations()


Regarding my comment on the special handling for kernel stubs in _create_declarations, I think probably you just get rid of this call and inline the relevant part of _create_declarations here.

Hi @arporter, can I just check what you want here. If I inline the _create_declarations function and remove it, it will need to be inlined into both invoke_declarations and stub_declarations. This will create some code duplication - are you happy with this cost to remove _create_declarations?

Hi Alistair, that's not quite what I meant. You can keep _create_declarations but move the stub-specific code out of it. There will be a little bit of duplication but I don't think it's much.

Ok sure, thanks for the clarification 😄

arporter · 2026-01-19T14:28:40Z

src/psyclone/tests/domain/lfric/arg_ordering_test.py

        create_arg_list.scalar(arg)
    const = LFRicConstants()
-    assert (f"Expected argument type to be one of {const.VALID_SCALAR_NAMES} "
+    assert (f"Expected argument type to be one of "


You can probably extend this test to get coverage of that missed line. You'll neeed to create a create_arg_list = KernCallArgList(kernel) and then call scalar on it with an argument that isn't a scalar. If this works, please update the name and docstring of the test.

mo-alistairp added 2 commits July 24, 2025 09:25

#1312: create initial ScalarArray example

bf523ba

#1312: start adding code generation for ScalarArrays

dc1ec4a

mo-alistairp self-assigned this Aug 19, 2025

mo-alistairp added in progress LFRic Issue relates to the LFRic domain labels Aug 19, 2025

Merge branch 'master' into 1312_scalararray_code_gen

862f368

arporter reviewed Aug 19, 2025

View reviewed changes

mo-alistairp added 8 commits August 19, 2025 13:54

#1312: lint fix

1a07148

#1312: condense the scalar_array functionality into the scalar function

decf6c1

#1312L lint fix

6d74811

#1312: fix the scalar type check logic

71c7d0c

#1312: remove the empty test file for now

fa6a7e6

#1312: fix the failing test

ae9ba03

#1312: actually fix the failing test

4961046

#1312: lint fix

bc8260f

mo-alistairp added 10 commits August 21, 2025 14:00

#1312: Add initial test implementation of ScalarArrays

8e2b3de

#1312: fix tests to check that gh_scalar_array is being assigned prop…

5b8c42d

…erly

1312: add TODO to check whether is_scalar references need updating

ea0e8f7

Merge branch 'master' into 1312_scalararray_code_gen

3b6ad7d

#1312: lint fix

33caeac

#1312: create a loose structure for acc access

605d9ab

#1312: Code gen and stub gen for ScalarArrays

c9a5598

#1312: lint fix

bbb37b7

Merge branch 'master' into 1312_scalararray_code_gen

9f3cfb5

#1312: lint fix

8e311c2

mo-alistairp added 3 commits October 22, 2025 15:33

#1312: lint fixes

ea6ae8d

#1312: move dimension array to lfric_scalar_args

136d89e

#1312: add declarations for ScalarArray

fd0d546

mo-alistairp added 8 commits December 9, 2025 16:47

#1312: Improve comments to distinguish between scal and decl_scal

58ba124

#1312: Include the dimensions array in the kernel call list

e9c7b77

#1312: remove old todo

3febeb7

#1312: Change the _array_ndims to be >=1

6af9a55

#1312: Add comment to _array_ndims

a7f3e98

#1312: change the test codegen to use get_invoke

9c8499d

#1312: add field related parameters to the testkern

49d89e5

Merge branch 'master' into 1312_scalararray_code_gen

5d92498

mo-alistairp mentioned this pull request Jan 12, 2026

Add GH_SCALAR_ARRAY to argument_mod MetOffice/lfric_core#213

Open

mo-alistairp added 4 commits January 15, 2026 17:23

#1312: Swap the order of dims_array and ScalarArray declarations

2741a4b

#1312: Update ScalarArray tests to match declaration order

444d99a

#1312: Add GH_SCALAR_ARRAY as an LFRic argument

b2d4553

#1312: lint fix

adcdb37

mo-alistairp requested a review from arporter January 16, 2026 12:30

mo-alistairp added ready for review and removed reviewed with actions labels Jan 16, 2026

Merge branch 'master' into 1312_scalararray_code_gen

d2a7ef9

arporter added under review and removed ready for review labels Jan 19, 2026

arporter mentioned this pull request Jan 19, 2026

PSyclone reordering symbol table during output resulting in compiler error #2202

Open

mo-alistairp mentioned this pull request Jan 19, 2026

Add GH_SCALAR_ARRAY to argument_mod MetOffice/lfric_core#223

Draft

28 tasks

arporter requested changes Jan 19, 2026

View reviewed changes

Merge branch 'master' into 1312_scalararray_code_gen

f187466

arporter temporarily deployed to integration January 19, 2026 14:35 — with GitHub Actions Inactive

arporter added enhancement reviewed with actions and removed under review labels Jan 19, 2026

#1312 ScalarArray code generation #3101

Are you sure you want to change the base?

#1312 ScalarArray code generation #3101

Conversation

mo-alistairp commented Aug 19, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

codecov bot commented Aug 19, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

mo-alistairp commented Oct 21, 2025

Uh oh!

mo-alistairp commented Dec 19, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

mo-alistairp commented Jan 16, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

mo-alistairp commented Jan 16, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

arporter 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!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

codecov bot commented Aug 19, 2025 •

edited

Loading

mo-alistairp commented Dec 19, 2025 •

edited

Loading

mo-alistairp commented Jan 16, 2026 •

edited

Loading

mo-alistairp commented Jan 16, 2026 •

edited

Loading