On Fri, Dec 09, 2022 at 10:18:34AM +0100, Martin Liška wrote:
> I'm going to push the revision.
>
> What exactly do you mean by random? I just know there are differences in
> between
> x86 and ppc:
>
> int __builtin_cpu_supports(const char *feature)
> This function returns a positive integer if the run-time CPU supports feature
> and returns 0 otherwise.
>
> This function returns a value of 1 if the run-time CPU supports the HWCAP
> feature feature and returns 0 otherwise.
Because x86-64-v2 etc. isn't a HWCAP feature, but rather an architecture
(or set of canned HWCAP features). So __builtin_cpu_is for those,
or especially for __builtin_cpu_is ("x86-64") would make more sense.
Though I see that many of the valid -march= values actually aren't supported
by __builtin_cpu_is either, whether it is pentium, nocona, lakemont etc.
but does support say some vendors (intel, amd).
Jakub
