Android and iOS promoted to Tier 3 platforms

--===============7124351967109415053==
Content-Type: multipart/alternative; boundary="0000000000000cd2850656d085c5"

--0000000000000cd2850656d085c5
Content-Type: text/plain; charset="UTF-8"

Hi all,

We just had two very long-running PRs merged to add source-level support
and CI jobs for Android and iOS. Kudos to Malcolm Smith and Russell
Keith-Magee for getting those PRs, and all the work needed for it in
CPython, PyPI, cibuildwheel, auditwheel, Meson, meson-python, pip, uv and
perhaps more tools in good shape. The NumPy PRs themselves are now pretty
straightforward, mostly test suite skips because these platforms don't
support subprocesses (just like Pyodide). For those who are interested in
the details:
https://github.com/numpy/numpy/pull/28759
https://github.com/numpy/numpy/pull/30412

I just opened a PR to document the Tier 3 promotion in NEP 57:
https://github.com/numpy/numpy/pull/32032.

Both platforms should support the Limited C API and Stable ABI just fine.
Experimental work in NumPy is ongoing to figure out if we can support that
at the source level (see https://github.com/numpy/numpy/issues/31913); for
mainstream platforms it's likely not okay performance-wise to ship only
Stable ABI wheels, but for newer/niche/legacy platforms, that could be a
solution to keep the wheel build load manageable. So when the people
supporting these platforms come asking for Tier 2 status after running in
CI without problems for a couple of months, we have a potential solution
that doesn't explode our wheel build matrix (the queue for these requests
is long: RISC-V, Pyodide, Android, iOS, ppc64le).

For right now, it's now straightforward to build numpy wheels from source
(or grab them off of a regular CI job as artifact), so if anyone is
interested in trying out NumPy on their phone, now is a good time to do so.

Cheers,
Ralf

--0000000000000cd2850656d085c5
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div>Hi all,</div><div><br></div><div>We just had two very=
 long-running PRs merged to add source-level support and CI jobs for Androi=
d and iOS. Kudos to Malcolm Smith and Russell Keith-Magee for getting those=
 PRs, and all the work needed for it in CPython, PyPI, cibuildwheel, auditw=
heel, Meson, meson-python, pip, uv and perhaps more tools in good shape. Th=
e NumPy PRs themselves are now pretty straightforward, mostly test suite sk=
ips=C2=A0because these platforms don&#39;t support subprocesses (just like =
Pyodide). For those who are interested in the details:</div><div><a href=3D=
"https://github.com/numpy/numpy/pull/28759">https://github.com/numpy/numpy/=
pull/28759</a></div><div><a href=3D"https://github.com/numpy/numpy/pull/304=
12">https://github.com/numpy/numpy/pull/30412</a></div><div><br></div><div>=
I just opened a PR to document the Tier 3 promotion in NEP 57:=C2=A0<a href=
=3D"https://github.com/numpy/numpy/pull/32032">https://github.com/numpy/num=
py/pull/32032</a>.</div><div><br></div><div>Both platforms should support t=
he Limited C API and Stable ABI just fine. Experimental work in NumPy is on=
going to figure out if we can support that at the source level (see=C2=A0<a=
 href=3D"https://github.com/numpy/numpy/issues/31913">https://github.com/nu=
mpy/numpy/issues/31913</a>); for mainstream platforms it&#39;s likely not o=
kay performance-wise=C2=A0to ship only Stable ABI wheels, but for newer/nic=
he/legacy platforms, that could be a solution to keep the wheel build load =
manageable. So when the people supporting these platforms come asking for T=
ier 2 status after running in CI without problems for a couple of months, w=
e have a potential solution that doesn&#39;t explode our wheel build matrix=
 (the queue for these requests is long: RISC-V, Pyodide, Android, iOS, ppc6=
4le).</div><div><br></div><div>For right now, it&#39;s now straightforward =
to build numpy wheels from source (or grab them off of a regular CI job as =
artifact), so if anyone is interested in trying out NumPy on their phone, n=
ow is a good time to do so.</div><div><br></div><div>Cheers,<br></div><div>=
Ralf</div><div><br></div></div>

--0000000000000cd2850656d085c5--

--===============7124351967109415053==
Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

_______________________________________________
NumPy-Discussion mailing list -- [email protected]
To unsubscribe send an email to [email protected]
https://mail.python.org/mailman3//lists/numpy-discussion.python.org
Member address: [email protected]

--===============7124351967109415053==--