From: Klemens Nanni <kn@openbsd.org>
Subject: Re: Numpy segfaults on amd64
To: ports@openbsd.org, Theo Buehler <tb@theobuehler.org>, Stuart Cassoff <exokoide@yahoo.ca>
Cc: Stuart Henderson <stu@spacehopper.org>, Daniel Dickman <didickman@gmail.com>, OpenBSD Ports <ports@openbsd.org>
Date: Sat, 17 Jan 2026 10:10:43 +0000

16 января 2026 г. 22:23:02 UTC, Theo Buehler <tb@theobuehler.org> пишет:
>> CPU Optimization Options
>>   baseline:
>>     Requested : min
>>     Enabled   : X86_V2
>>   dispatch:
>>     Requested : max
>>     Enabled   : X86_V3 X86_V4 AVX512_ICL AVX512_SPR
>
>With the below diff I get this:
>
>CPU Optimization Options
>  baseline:
>    Requested : none
>    Enabled   :
>  dispatch:
>    Requested : max
>    Enabled   : X86_V2 X86_V3 X86_V4 AVX512_ICL AVX512_SPR
>
>If it works, I think it is what we want. This is now unmaintained and
>may break at some point, but for now I guess it's the best we can do:
>keep the old boxes working, newer hardware gets to use the fancy code.
>
>https://github.com/numpy/numpy/blob/729d61c84d087dccf17e4ef960391227aaa0c5b3/doc/source/reference/simd/build-options.rst#targeting-older-cpus

OK kn

>
>Index: Makefile
>===================================================================
>RCS file: /cvs/ports/math/py-numpy/Makefile,v
>diff -u -p -r1.96 Makefile
>--- Makefile	12 Jan 2026 05:39:02 -0000	1.96
>+++ Makefile	16 Jan 2026 22:16:39 -0000
>@@ -4,6 +4,8 @@ MODPY_DISTV=	2.4.1
> DISTNAME=	numpy-${MODPY_DISTV}
> PKGNAME=	py-${DISTNAME}
> 
>+REVISION=	0
>+
> CATEGORIES=	math devel
> 
> HOMEPAGE=	https://numpy.org/
>@@ -78,6 +80,9 @@ pre-build:
> # https://github.com/numpy/numpy/issues/20799
> .if ${MACHINE_ARCH} == i386
> 	sed -i '/^option.*disable-optim/ s/false/true/' ${WRKSRC}/meson.options
>+.endif
>+.if ${MACHINE_ARCH} == amd64
>+MODPY_PYBUILD_ARGS += -Csetup-args=-Dcpu-baseline="none"
> .endif
> 
> do-test:
>