Fix identity values in global reductions

[msimberg]

Makes sure the correct datatype is selected for the identity values when a rank has no local data. Also chooses inf/-inf for floats and (finite) min/max values for integral types.

Also adds one more test input with integral values.

[Copilot]

Pull request overview

This PR fixes identity values used in global reductions when a rank has no local data. The changes ensure proper datatype selection by using infinity for floating-point types and min/max integer limits for integral types. Additionally, a new test case with integer values is added to verify the fix.

Changes:

• Added new static helper methods to compute appropriate identity values for min, max, and sum operations based on the data type
• Updated global reduction methods to use these type-aware identity values instead of hardcoded values
• Added integer test case [-10, 20, 4] and updated type hints to support both float and integer inputs

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
model/common/src/icon4py/model/common/decomposition/mpi_decomposition.py	Adds three static identity methods and updates min/max/sum/mean to use dtype-specific identity values
model/common/tests/common/decomposition/mpi_tests/test_mpi_decomposition.py	Adds integer test input and updates type hints from list[float] to list[data_alloc.ScalarT]

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[msimberg]

@nfarabullini would you mind being manual CI for this PR, and checking that the tests pass for you locally as well? I've tested this locally where it passes (but the tests passed previously as well). I've also added the changes to #692 where the tests used to fail but now pass.

[nfarabullini]

cscs-ci run default

[msimberg]

cscs-ci run default

[github-actions]

Mandatory Tests

Please make sure you run these tests via comment before you merge!

• cscs-ci run default

Optional Tests

To run benchmarks you can use:

• cscs-ci run benchmark-bencher

To run tests and benchmarks with the DaCe backend you can use:

• cscs-ci run dace

To run test levels ignored by the default test suite (mostly simple datatest for static fields computations) you can use:

• cscs-ci run extra

For more detailed information please look at CI in the EXCLAIM universe.

[msimberg]

cscs-ci run default