Re: ENH: Add float16, npy_half, specific routines to the NumPy C-API as NpyHalf*

Francesc Alted via NumPy-Discussion <[email protected]> Mon, 24 Nov 2025 12:24:15 +0100
Newsgroups gmane.comp.python.numeric.general
Message-ID <CAFrp1vpVnyXG37E9A5unAW2-86qxyZV0vnp2SLRFiR+zpyFZCw@mail.gmail.com>
On Mon, Nov 24, 2025 at 11:30 AM Ilhan Polat via NumPy-Discussion <
[email protected]> wrote:

> Whatever the decision would be, could we at least try to use the bit width
> in the name such as float16 as much as possible? Half means nothing and
> gives different platforms different ideas which would be very annoying
> after a decade, every platform went their own way. Also syntactically
> double -> float -> half (not to mention the "long" situation) is not
> consistent either. There is still no guarantee obviously (say complex256
> support on windows) but at least we eliminate the obvious ones.
>

+1

-- 
Francesc Alted

_______________________________________________
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]