Deprecating mixed precision inner product subroutines

[Matthew Walker (mattatmet)]

PR Summary

Sci/Tech Reviewer: thomasmelvin
Code Reviewer: Oakley Brunt (@oakleybrunt)

This is a PR reflecting the Trac ticket #4449. The changes in this PR are a like-for-like replica of that ticket. This involves removing the PSyKAL light code introduced to perform mixed precision inner product operations in order to deprecate that feature. These inner product operations have been replaced in the solver with PSyclone built-ins that were introduced in PSyclone#1786.

• linked Deprecating mixed precision inner product and kgo changes lfric_apps#501

Code Quality Checklist

• I have performed a self-review of my own code
• My code follows the project's style guidelines
• Comments have been included that aid understanding and enhance the readability of the code
• My changes generate no new warnings
• All automated checks in the CI pipeline have completed successfully

Testing

• I have tested this change locally, using the LFRic Core rose-stem suite
• If required (e.g. API changes) I have also run the LFRic Apps test suite using this branch
• If any tests fail (rose-stem or CI) the reason is understood and acceptable (e.g. kgo changes)
• I have added tests to cover new functionality as appropriate (e.g. system tests, unit tests, etc.)
• Any new tests have been assigned an appropriate amount of compute resource and have been allocated to an appropriate testing group (i.e. the developer tests are for jobs which use a small amount of compute resource and complete in a matter of minutes)

trac.log

Test Suite Results - lfric_core - mixed-precision-core/run1

Suite Information

Item	Value
Suite Name	mixed-precision-core/run1
Suite User	matthew.walker
Workflow Start	2026-05-19T09:23:57
Groups Run	all

Dependency	Reference	Main Like
lfric_core	mattatmet/lfric_core@mixed-precision	False
SimSys_Scripts	MetOffice/SimSys_Scripts@2025.12.1	True

Task Information

❌ failed tasks - 1

Task	State
validate_rose_meta	failed

✅ succeeded tasks - 382

⌛ waiting tasks - 1

Task	State
housekeep_azspice	waiting

Security Considerations

• I have reviewed my changes for potential security issues
• Sensitive data is properly handled (if applicable)
• Authentication and authorisation are properly implemented (if applicable)

Performance Impact

• Performance of the code has been considered and, if applicable, suitable performance measurements have been conducted

AI Assistance and Attribution

• Some of the content of this change has been produced with the assistance of Generative AI tool name (e.g., Met Office Github Copilot Enterprise, Github Copilot Personal, ChatGPT GPT-4, etc) and I have followed the Simulation Systems AI policy (including attribution labels)

Documentation

• Where appropriate I have updated documentation related to this change and confirmed that it builds correctly

PSyclone Approval

• If you have edited any PSyclone-related code (e.g. PSyKAl-lite, Kernel interface, optimisation scripts, LFRic data structure code) then please contact the TCD Team

Sci/Tech Review

• I understand this area of code and the changes being added
• The proposed changes correspond to the pull request description
• Documentation is sufficient (do documentation papers need updating)
• Sufficient testing has been completed

(Please alert the code reviewer via a tag when you have approved the SR)

Code Review

• All dependencies have been resolved
• Related Issues have been properly linked and addressed
• CLA compliance has been confirmed
• Code quality standards have been met
• Tests are adequate and have passed
• Documentation is complete and accurate
• Security considerations have been addressed
• Performance impact is acceptable

[Alistair Pirrie (mo-alistairp)]

Happy with these changes from a TCD team/PSyclone perspective - thanks for removing some PSyKAl-lite!

[Thomas Bendall (tommbendall)]

Always good to see some psykal-light code removed. I don't think the corresponding innerproduct routines are called in LFRic-Apps so no linked PR is necessary?

If you haven't already, it might be worth double-checking by running an LFRic-Apps branch linked to this Core branch.

[Matthew Walker (mattatmet)]

Always good to see some psykal-light code removed. I don't think the corresponding innerproduct routines are called in LFRic-Apps so no linked PR is necessary?

If you haven't already, it might be worth double-checking by running an LFRic-Apps branch linked to this Core branch.

Thanks for the quick review Tom (Thomas Bendall (@tommbendall))! There's actually an integration test in apps that has these inner product calls that I've had to change as well as a significant number of KGOs that have to be adjusted as well. I've put those into #501 in apps. So I think it might be necessary but happy to be wrong on that?

[Thomas Bendall (tommbendall)]

Always good to see some psykal-light code removed. I don't think the corresponding innerproduct routines are called in LFRic-Apps so no linked PR is necessary?
If you haven't already, it might be worth double-checking by running an LFRic-Apps branch linked to this Core branch.

Thanks for the quick review Tom (Thomas Bendall (Thomas Bendall (@tommbendall)))! There's actually an integration test in apps that has these inner product calls that I've had to change as well as a significant number of KGOs that have to be adjusted as well. I've put those into #501 in apps. So I think it might be necessary but happy to be wrong on that?

Ah sorry, I completely missed that there already was a linked Apps PR! So I'm not sure that I understand why every KGO changes... are the sums now happening at the native precision rather than r_double precision?

[Matthew Walker (mattatmet)]

Ah sorry, I completely missed that there already was a linked Apps PR! So I'm not sure that I understand why every KGO changes... are the sums now happening at the native precision rather than r_double precision?

Thomas Bendall (@tommbendall), yes, the psykal-lite code promotes the solver field within the subroutine to r_double, whereas the PSyclone built-ins (I think) are set to use whatever the precision is of the argument that it's given, which will be r_solver.

[Thomas Bendall (tommbendall)]

Ah sorry, I completely missed that there already was a linked Apps PR! So I'm not sure that I understand why every KGO changes... are the sums now happening at the native precision rather than r_double precision?

Thomas Bendall (Thomas Bendall (@tommbendall)), yes, the psykal-lite code promotes the solver field within the subroutine to r_double, whereas the PSyclone built-ins (I think) are set to use whatever the precision is of the argument that it's given, which will be r_solver.

I thought that must be the case! I have chatted with Tom M and Chris M and am happy that this is fine -- I also ran some NWP case studies to check that the impact is negligible and stability wouldn't be impacted, and everything looked OK.

[Matthew Walker (mattatmet)]

I thought that must be the case! I have chatted with Tom M and Chris M and am happy that this is fine -- I also ran some NWP case studies to check that the impact is negligible and stability wouldn't be impacted, and everything looked OK.

Awesome, thanks for doing that! Glad they were stable.