Re: Floating point on 32-bit x86: x87 precision (XPFPA) and fpmath=sse
| From: | marc at mabe dot berlin | Date: | Wed, 11 Nov 2026 08:17:55 +0000 |
| Subject: | Re: Floating point on 32-bit x86: x87 precision (XPFPA) and fpmath=sse | ||
| References: | 1 2 | Groups: | php.internals |
| Request: | Send a blank email to internals+get-133068@lists.php.net to get a copy of this message | ||
> On 8.11.2026 15:25 CET Sjoerd Langkemper <sjoerd-php@linuxonly.nl> wrote:
>
>
> On Sun, Nov 8, 2026, at 14:17, marc@mabe.berlin
> mailto:marc@mabe.berlin wrote:
>
> > Can we enable
-msse2 -mfpmath=sse by default on x86
> > (require SSE2).
> >
>
> Yes. As I understand it we are not really interested in supporting actual 32 bit hardware, but
> running 32 bit builds on WASM or to save memory. So breaking compatibility with old CPUs seems fair.
>
I have opened a PR to enable SSE2 and fpmath=sse on x86 by default without configure option. https://github.com/php/php-src/pull/25362
>
> I think Windows already requires SSE2?
>
Windows builds should use SSE2 via --enable-native-intrinsics=sse2 and should enable it
for double arithmetic. I do not have windows to verify this.
>
> Regards,
>
> Sjoerd Langkemper
>