[Bug tree-optimization/66917] [4.9/5/6 regression] ARM: NEON: memcpy compiles to vld1 and vst1 with incorrect alignment

["ktkachov at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66917

--- Comment #8 from ktkachov at gcc dot gnu.org ---
The 007t.lower one looks like:
  try
    {
      MEM[(char * {ref-all})&a] = MEM[(char * {ref-all})ap];
      MEM[(char * {ref-all})&b] = MEM[(char * {ref-all})bp];
      D.5587 = a.u[0];
      D.5588 = b.u[0];
      D.5589 = D.5587 ^ D.5588;
      a.u[0] = D.5589;
      D.5590 = a.u[1];
      D.5591 = b.u[1];
      D.5592 = D.5590 ^ D.5591;
      a.u[1] = D.5592;
      MEM[(char * {ref-all})outp] = MEM[(char * {ref-all})&a];
    }
  finally
    {
      a = {CLOBBER};
      b = {CLOBBER};
    }
  return;