# Numba Meeting: 2025-06-24 Attendees: Todd, Graham, Val, Siu, Guilherme, Stuart, Kaustubh FPOC (last week): Todd FPOC (incoming): Val NOTE: All communication is subject to the [Numba Code of Conduct](https://github.com/numba/numba-governance/blob/accepted/code-of-conduct.md). Please refer to [this calendar](https://calendar.google.com/calendar/embed?src=5rqnddm4gjrdfjm31fsv182epk%40group.calendar.google.com) for the next meeting date. ## ~~Office hour signup~~ ## Notes - Get rid of make in the build systems, switching everything to CMAKE. - Default to static LLVM linking. ## Discussion - When is LLVM 20 going to land in llvmlite ? - https://github.com/numba/llvmlite/pull/1226 - GHA llvmdev 20 build - Hopefully next release. - General opinion on the idea of _[RFC] Support for `__array__()`_ https://github.com/numba/numba/pull/9584 - Going to do a test where we have a type that we implement explicitly but also has `__array__` and then want to make sure that the explicit type support is the one that is used. - Should add this as a test cast - How would this effect awkward array? - Adding protocol-based fingerprinting ahead of typeof implementations will break any existing code that expects to be able to type things that support the array function protocol using a typeof implementation. - Therefore, it's probably better to leave this implementation largely as-is as a last resort - This prefers correctness over performance, which is the right choice for now - Things we will not consider now, but may consider later: - Further performance improvement - User-definined protocol-based typing extension - Next steps: - Fix up the suggestions on the PR regarding checking return values - Add comments or something documenting how the protocol-based array function typing works. - Add the test case as mentioned above - Discuss admission of `INSERT NAME HERE` to numba-maintainers and this triage meeting. ## New "Ready for Review" PRs ## 1. New Issues - [numba#10123](https://github.com/numba/numba/issues/10123) - Numba exhibits strange behavior when doing bitwise operations ### Closed Issues ## 2. New PRs - [numba#10122](https://github.com/numba/numba/pull/10122) - add ufunc.reduceat support - [numba#10124](https://github.com/numba/numba/pull/10124) - Implementing support for atomics in CPU numba - Todd will leave them some comments. - [numba#10125](https://github.com/numba/numba/pull/10125) - #9977 rebased - [llvmlite#1234](https://github.com/numba/llvmlite/pull/1234) - Default to static linking, override with LLVMLITE_SHARED - [llvmlite#1235](https://github.com/numba/llvmlite/pull/1235) - Update pre-commit hook PyCQA/flake8 to v7.3.0 - [llvmlite#1236](https://github.com/numba/llvmlite/pull/1236) - Update pre-commit hook python-jsonschema/check-jsonschema to v0.33.1 ### Closed PRs - [numba#10120](https://github.com/numba/numba/pull/10120) - Bug: Reproducer for platform dependent compilation issue with OS X - [numba#10121](https://github.com/numba/numba/pull/10121) - test/numba win wmic (last numba: 10125; llvmlite 1236)