https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70729
--- Comment #32 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Also, the testcase has weird dg- directives:
// { dg-do compile }
// { dg-require-effective-target vect_simd_clones }
// { dg-additional-options "-Ofast" }
// { dg-additional-options "-mavx2 -fopenmp-simd" { target x86_64-*-* i?86-*-*
} }
It doesn't use declare simd, so the dg-require-effective-target makes no sense.
-fopenmp-simd can be applied on all targets, -mavx2 of course not, but then the
dg-final should be probably guarded with x86_64* / i?86* or whatever other
arches are known to vectorize it.
