numpy/numpy-user-dtypes

Test little and big endian architectures in CI

Closed this issue · 7 comments

We test little endian architectures in the numpy CI using PPC64le QEMU emulation. It's slow but it at least lets you do a real end-to-end test.

Originally posted by @ngoldbaum in #118 (comment)

@SwayamInSync Maybe you have an idea for how to set this up? Once the infrastructure is there, I could then implement the signbit ufunc with its binary implementation (and follow that up with nextafter).

"PPC64le" is also little-endian (we are compatible & well tested on little endian), I can try emulating the Big-Endian variants like PPC64 and s390x with QEMU

Ah yeah, sorry, got it backwards!

@SwayamInSync Do you have an update on this?

@SwayamInSync Do you have an update on this?

Yup, I started the work on this in a local fork, its just QEMU emulation it takes ~1.5 hrs to build the image and test. Will drop the PR as it gets done

Maybe setting up ccache on github CI will help a little? At least after the first build.

@SwayamInSync How's your work on the QEMU PR coming along?