RE: [PATCH 2/5] aarch64: Added optimized memcpy and memmove for A64FX

["naohirot@fujitsu.com" <naohirot@fujitsu.com>]

Szabolcs-san,

Thank you for your review.

> >  /* Maximum number of IFUNC implementations.  */
> > -#define MAX_IFUNC	4
> > +#if HAVE_SVE_ASM_SUPPORT
> > +# define MAX_IFUNC	7
> > +#else
> > +# define MAX_IFUNC	6
> > +#endif
> 
> hm this MAX_IFUNC looks a bit problematic: currently its only use is to detect if a
> target requires more ifuncs than the array passed to __libc_ifunc_impl_list, but for
> that ideally it would be automatic, not manually maintained.
> 
> i would just define it to 7 unconditionally (the maximum over valid configurations).

OK, I'll fix it to 7 unconditionally.

> > cores diff --git a/sysdeps/aarch64/multiarch/init-arch.h
> > b/sysdeps/aarch64/multiarch/init-arch.h
> > index a167699e74..d20e7e1b8e 100644
> > --- a/sysdeps/aarch64/multiarch/init-arch.h
> > +++ b/sysdeps/aarch64/multiarch/init-arch.h
> > @@ -33,4 +33,6 @@
> >    bool __attribute__((unused)) bti =
> \
> >      HAVE_AARCH64_BTI && GLRO(dl_aarch64_cpu_features).bti;
> 	      \
> >    bool __attribute__((unused)) mte =
> \
> > -    MTE_ENABLED ();
> > +    MTE_ENABLED ();
> 	      \
> > +  unsigned __attribute__((unused)) sve =
> \
> > +    GLRO(dl_aarch64_cpu_features).sve;
> 
> i would use bool here.

I'll fix it to the bool.

> > --- /dev/null
> > +++ b/sysdeps/aarch64/multiarch/memcpy_a64fx.S
> 
> skipping this.

I wait for your review.

Thanks.
Naohiro