This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: V6 [PATCH] x86: Add <sys/platform/x86.h>
- From: Florian Weimer <fweimer at redhat dot com>
- To: "H.J. Lu" <hjl dot tools at gmail dot com>
- Cc: libc-alpha at sourceware dot org
- Date: Fri, 26 Oct 2018 06:53:37 +0200
- Subject: Re: V6 [PATCH] x86: Add <sys/platform/x86.h>
- References: <CAMe9rOqcfBAsSQQme3VJyMKsbAvKO_YdaG=1ZJqrm+T8aPztag@mail.gmail.com> <87efcekula.fsf@oldenburg.str.redhat.com> <CAMe9rOp_jL00S63ou9kiGiLP3spSPASGv-2s5v1zOagVf=JbUA@mail.gmail.com>
* H. J. Lu:
> On 10/25/18, Florian Weimer <fweimer@redhat.com> wrote:
>> * H. J. Lu:
>>
>>
>>> Here is the patch with CPU_FEATURE_USABLE only. OK for master
>>> branch?
>>
>> Thanks. Recently, the matter of special-casing code inside target
>> clones came up on the glibc list:
>
> I am going to check it in.
>
>> <https://gcc.gnu.org/ml/gcc-help/2018-10/msg00118.html>
>>
>> Do you think it might be possible to eventually turn some results for
>> CPU_FEATURE_USABLE in target clones into compile-time constants because
>> the compiler knows the results of the feature selection in the
>> automatically generated IFUNC resolver?
>
> By definition, CPU_FEATURE_USABLE is evaluated at run-time.
> Can GCC be extended to allow intrinsics in target_clone.
I am wondering if we can tweak CPU_FEATURE_USABLE in such a way that it
can be easily compatible with future intrinsics.
Thanks,
Florian