Re: A modest proposal

--===============5053602109514015936==
Content-Type: multipart/alternative; boundary="0000000000006b943806519a97df"

--0000000000006b943806519a97df
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Tue, May 12, 2026 at 9:32=E2=80=AFAM Matthew Brett via NumPy-Discussion =
<
[email protected]> wrote:

> Hi,
>
> On Tue, May 12, 2026 at 6:52=E2=80=AFAM matti picus via NumPy-Discussion
> <[email protected]> wrote:
> >
> > On Tue, May 12, 2026 at 3:42=E2=80=AFAM Charles R Harris via NumPy-Disc=
ussion
> > <[email protected]> wrote:
> > >
> > > Hi All,
> > >
> > > I would like to propose dropping 32 bit windows support in the NumPy
> 2.6 branch. It is time, Steam is dropping support, and most downstream
> projects dropped support some time ago. We can still test it, perhaps, bu=
t
> the world isn't going that way.
>

Keeping it in CI for now while dropping the wheels for 2.6.0 seems like a
good idea to me. Keeping our wheel build matrix reasonable is important for
maintainability, and 32-bit Python on Windows has very low usage numbers
now (and almost all of those are simple mistakes of clicking the wrong
installer link on python.org).

There have also been no serious complaints when other large packages like
SciPy, Pillow and Matplotlib dropped support.

For completeness, NumPy's download numbers were <1% years ago already, and
dropping off further with every new Python version:
https://discuss.python.org/t/dropping-32-bit-packages/5476/13. And we've
been shipping 32-bit wheels without BLAS for years, and to my knowledge
exactly zero people have complained about linalg operations being ~100x
slower. So, it's time.

Now if we keep one CI job, does anyone care enough about that to list
themselves behind that job in
https://numpy.org/neps/nep-0057-numpy-platform-support.html#tier-3? If not,
we can put it in Unsupported perhaps, with a note that we kept one CI job
that can be dropped as soon as it becomes more trouble than it's worth.


> > >
> > > Chuck
> >
> > +1 from me
> > Matti
>
> Worth floating with Steve Dower?  https://stevedower.id.au/about ;
> https://github.com/zooba


Here's Steve's recommendation at the end of the most recent last thread on
Discourse about 32-bit Windows (from
https://discuss.python.org/t/consider-downgrading-windows-32-bit-from-tier-=
1-to-tier-2-or-tier-3-in-python-3-13/33719/63
):

"If package maintainers want to stop releasing wheels for 32-bit Windows,
that=E2=80=99s between them and their users. Right now, dropping 32-bit Win=
dows and
picking up ARM64 Windows is probably a good trade, though in most cases the
cost will be close to zero anyway."

Cheer,s
Ralf



>
>
> I seem to remember him saying, quite a few years ago, that 32-bit
> Windows was in practice only important for Windows ARM users running
> the 32-bit Intel binaries under emulation.  Now there are Windows ARM
> wheels (and CI), so I bet that there are a very small number of people
> who need these.
>
> Cheers,
>
> Matthew
> _______________________________________________
> 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]
>

--0000000000006b943806519a97df
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div dir=3D"ltr"><br></div><br><div class=3D"gmail_quote g=
mail_quote_container"><div dir=3D"ltr" class=3D"gmail_attr">On Tue, May 12,=
 2026 at 9:32=E2=80=AFAM Matthew Brett via NumPy-Discussion &lt;<a href=3D"=
mailto:[email protected]">[email protected]</a>&gt; wro=
te:<br></div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px =
0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Hi,<br>
<br>
On Tue, May 12, 2026 at 6:52=E2=80=AFAM matti picus via NumPy-Discussion<br=
>
&lt;<a href=3D"mailto:[email protected]" target=3D"_blank">numpy-=
[email protected]</a>&gt; wrote:<br>
&gt;<br>
&gt; On Tue, May 12, 2026 at 3:42=E2=80=AFAM Charles R Harris via NumPy-Dis=
cussion<br>
&gt; &lt;<a href=3D"mailto:[email protected]" target=3D"_blank">n=
[email protected]</a>&gt; wrote:<br>
&gt; &gt;<br>
&gt; &gt; Hi All,<br>
&gt; &gt;<br>
&gt; &gt; I would like to propose dropping 32 bit windows support in the Nu=
mPy 2.6 branch. It is time, Steam is dropping support, and most downstream =
projects dropped support some time ago. We can still test it, perhaps, but =
the world isn&#39;t going that way.<br></blockquote><div><br></div><div>Kee=
ping it in CI for now while dropping the wheels for 2.6.0 seems like a good=
 idea to me. Keeping our wheel build matrix reasonable is important for mai=
ntainability, and 32-bit Python on Windows has very low usage numbers now (=
and almost all of those are simple mistakes of clicking the wrong installer=
 link on <a href=3D"http://python.org">python.org</a>).=C2=A0</div><div><br=
></div><div>There have also been no serious complaints when other large pac=
kages like SciPy, Pillow and Matplotlib dropped support.</div><div><br></di=
v><div>For completeness, NumPy&#39;s download numbers were &lt;1% years ago=
 already, and dropping off further with every new Python version:=C2=A0<a h=
ref=3D"https://discuss.python.org/t/dropping-32-bit-packages/5476/13">https=
://discuss.python.org/t/dropping-32-bit-packages/5476/13</a>. And we&#39;ve=
 been shipping 32-bit wheels without BLAS for years, and to my knowledge ex=
actly zero people have complained about linalg operations being ~100x slowe=
r. So, it&#39;s time.</div><div><br></div><div>Now if we keep one CI job, d=
oes anyone care enough about that to list themselves behind that job in=C2=
=A0<a href=3D"https://numpy.org/neps/nep-0057-numpy-platform-support.html#t=
ier-3">https://numpy.org/neps/nep-0057-numpy-platform-support.html#tier-3</=
a>? If not, we can put it in Unsupported perhaps, with a note that we kept =
one CI job that can be dropped as soon as it becomes more trouble than it&#=
39;s worth.</div><div>=C2=A0</div><blockquote class=3D"gmail_quote" style=
=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding=
-left:1ex">
&gt; &gt;<br>
&gt; &gt; Chuck<br>
&gt;<br>
&gt; +1 from me<br>
&gt; Matti<br>
<br>
Worth floating with Steve Dower?=C2=A0 <a href=3D"https://stevedower.id.au/=
about" rel=3D"noreferrer" target=3D"_blank">https://stevedower.id.au/about<=
/a> ;<br>
<a href=3D"https://github.com/zooba" rel=3D"noreferrer" target=3D"_blank">h=
ttps://github.com/zooba</a></blockquote><div><br></div><div>Here&#39;s Stev=
e&#39;s recommendation at the end of the most recent last thread on Discour=
se about 32-bit Windows (from=C2=A0<a href=3D"https://discuss.python.org/t/=
consider-downgrading-windows-32-bit-from-tier-1-to-tier-2-or-tier-3-in-pyth=
on-3-13/33719/63">https://discuss.python.org/t/consider-downgrading-windows=
-32-bit-from-tier-1-to-tier-2-or-tier-3-in-python-3-13/33719/63</a>):</div>=
<div><br></div><div>&quot;If package maintainers want to stop releasing whe=
els for 32-bit Windows, that=E2=80=99s between them and their users. Right =
now, dropping 32-bit Windows and picking up ARM64 Windows is probably a goo=
d trade, though in most cases the cost will be close to zero anyway.&quot;<=
/div><div><br></div><div>Cheer,s<br></div><div>Ralf</div><div><br></div><di=
v>=C2=A0</div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px=
 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><br>
<br>
I seem to remember him saying, quite a few years ago, that 32-bit<br>
Windows was in practice only important for Windows ARM users running<br>
the 32-bit Intel binaries under emulation.=C2=A0 Now there are Windows ARM<=
br>
wheels (and CI), so I bet that there are a very small number of people<br>
who need these.<br>
<br>
Cheers,<br>
<br>
Matthew<br>
_______________________________________________<br>
NumPy-Discussion mailing list -- <a href=3D"mailto:numpy-discussion@python.=
org" target=3D"_blank">[email protected]</a><br>
To unsubscribe send an email to <a href=3D"mailto:numpy-discussion-leave@py=
thon.org" target=3D"_blank">[email protected]</a><br>
<a href=3D"https://mail.python.org/mailman3//lists/numpy-discussion.python.=
org" rel=3D"noreferrer" target=3D"_blank">https://mail.python.org/mailman3/=
/lists/numpy-discussion.python.org</a><br>
Member address: <a href=3D"mailto:[email protected]" target=3D"_b=
lank">[email protected]</a><br>
</blockquote></div></div>

--0000000000006b943806519a97df--

--===============5053602109514015936==
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]

--===============5053602109514015936==--