Christopher Sidebottom

Ok, so I played a bit of spot the difference and left some comments, and I ran some benchmarks quickly - it looks like with ASIMD, there's regressions with these:...

@jan-wassenberg just looking at the CI failures and it seems there's some compiler incompatibility on PPC (https://github.com/numpy/numpy/actions/runs/8087779139/job/22100478961?pr=25781#step:8:504) and an abort on Z13 (https://github.com/numpy/numpy/actions/runs/8087779139/job/22100479504?pr=25781) any ideas? There's also a failure on...

Looks like the new AArch64 CI failure is due to missing `per_target.cc` but we really shouldn't need it if we remove the dependency on `targets.cc` - will rebase https://github.com/google/highway/pull/1935 and...

https://github.com/google/highway/pull/2063 should fix the `armhf` failures, at least in my qemu 😸

@r-devulap you should be able to change this line to `abort.cc`: https://github.com/numpy/numpy/blob/0be3c26962838fb5e5fac15c3b7600f7f69f57c2/numpy/_core/meson.build#L108 That should remove all the dynamic dispatch code and fix the symbol not found error on AArch64 😸

> FYI you can set `HWY_BASELINE_TARGETS` to your desired target e.g., (HWY_AVX3|HWY_AVX2|HWY_SSE4|HWY_EMU128), to override Highway's detection based on compiler flags. At the moment, the constants are generated by Meson here:...

> The benefits of using the updated Accelerate library are (1) performance (xref [gh-24053](https://github.com/numpy/numpy/pull/24053) for benchmarks) and (2) smaller wheels because we're no longer vendoring OpenBLAS. Hi @rgommers, is the...

@martin-frbg my concern with that is that the source of truth is no longer `param.h`?

@martin-frbg , this is closer to what I was thinking previously, what do you think? I can see others have done similar in `setparam-ref.c` but I'm assuming that doesn't work...

Thanks @martin-frbg, I'll look into it 😸 !