Re: fixing AC_OPENMP for Apple clang

Dima Pasechnik <[email protected]> Mon, 02 Mar 2026 03:52:58 -0600
Newsgroups gmane.comp.sysutils.autoconf.general
Message-ID <[email protected]>
On March 2, 2026 2:49:53 AM CST, Florian Weimer <fweimer@redhat=2Ecom> wrot=
e:
>* Dima Pasechnik:
>
>> Since a while, Apple clang allows OpenMP (from llvm project, GNU OpenMP
>> doesn't work) if one supplies an option "-Xpreprocessor -fopenmp"=2E
>> The (unprefixed) option "-fopenmp" is not understood, and leads to an
>> error=2E
>
>Is this actually documented as supported by Apple?  I don't think so;
>this doesn't show any results: <https://developer=2Eapple=2Ecom/search/?q=
=3Dopenmp>
>
>I'm not sure it's a good idea to automatically enable compiler flags
>just because they do not immediately cause compilation to fail=2E
>Presumably there's a reason why this feature is somewhat hidden in the
>Apple compiler=2E

It is not unusual for commercial vendors to leave features undocumented=2E=
 They say that one of  the reasons behind is that Apple wants developers to=
 use their own parallelism tool called Grand Unified Dispatch=2E

The use of this flag is documented by a number of opensource projects, e=
=2Eg=2E by R Project=2E Meson supports this out of the box,
in the sense that you can specify OpenMP support in their meson=2Ebuild co=
nfig files, and it will enable OpenMP for your build with Apple Clang=2E Ho=
mebrew supports this=2E Etc etc=2E

Hope this explains=2E
Dima

>
>Thanks,
>Florian
>