Raghuveer Devulapalli

@ArielHeleneto could you fix the failing lint tests please?

Don't think so. Any idea why the test fails? Seems unrelated.

With [NEP 54](https://numpy.org/neps/nep-0054-simd-cpp-highway.html), NumPy decided to move to google highway, which is a well-established open source solution to this problem of using platform agnostic intrinsics. Maintaining our own version of...

> Since we rely on numpy to build astropy (and e.g. build our own ufuncs), I am not sure how astropy does this. Do you have the entire numpy source...

Highway isn't exposed through Numpy headers. If the goal here is to leverage SIMD for astropy core functions across multiple architectures, won't adding it as a submodule the easiest way...

If you goal is to determine if SIMD benefits your key benchmarks, I would start by prototyping with native intrinsics on your available hardware. Don't worry about portability or dynamic...

close/reopen to rerun the CI

@game-difficulty could you not just add `__attribute__((visibility("default")))` to the partition function and build the shared library? That should expose partition function per your requirement. EDIT: fixed incorrect attribute :)

Have you tried building with this line removed: https://github.com/intel/x86-simd-sort/blob/87486d17a7d916f2992800b0219e28d0af29e765/meson.build#L42

Hi @fabiocannizzo, I had a quick question: is the purpose of this PR to build this library on Windows? If that’s the case, do we also need to expose all...