https://gcc.gnu.org/bugzilla/show_bug.cgi?id=123484

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
-Ofast -m32 -mavx512vl -mavx512fp16 is enough to reproduce.
typedef __attribute__((__vector_size__ (8))) _Float16 F;
int a, b;
F f, *p;

void
foo ()
{
  do
    {
      f /= b ? *p : (F) {40};
l:
    }
  while (a);
  goto l;
}

Note, mulv4hf3 pattern isn't enabled in this case, because TARGET_MMX_WITH_SSE
is false.

Reply via email to