[Bug tree-optimization/66142] Loop is not vectorized because not sufficient support for GOMP_SIMD_LANE

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

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66142

--- Comment #7 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Slightly improved variant of #c6 testcase:
struct A { float x, y; };
struct B { struct A u; };
void bar (struct A *);

float
f1 (struct B *x, int y)
{
  struct A p;
  p.x = 1.0f;
  p.y = 2.0f;
  struct A *q = &x[y].u;
  *q = p;
  float f = x[y].u.x + x[y].u.y;
  bar (&p);
  return f;
}

float
f2 (struct B *x, int y)
{
  struct A p;
  p.x = 1.0f;
  p.y = 2.0f;
  x[y].u = p;
  float f = x[y].u.x + x[y].u.y;
  bar (&p);
  return f;
}

float
f3 (struct B *x, int y)
{
  struct A p;
  p.x = 1.0f;
  p.y = 2.0f;
  struct A *q = &x[y].u;
  __builtin_memcpy (&q->x, &p.x, sizeof (float));
  __builtin_memcpy (&q->y, &p.y, sizeof (float));
  *q = p;
  float f = x[y].u.x + x[y].u.y;
  bar (&p);
  return f;
}

float
f4 (struct B *x, int y)
{
  struct A p;
  p.x = 1.0f;
  p.y = 2.0f;
  __builtin_memcpy (&x[y].u.x, &p.x, sizeof (float));
  __builtin_memcpy (&x[y].u.y, &p.y, sizeof (float));
  float f = x[y].u.x + x[y].u.y;
  bar (&p);
  return f;
}