[Bug tree-optimization/109119] New: missing loading fre with branch comparing two pointers

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

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109119

            Bug ID: 109119
           Summary: missing loading fre with branch comparing two pointers
           Product: gcc
           Version: 13.0
            Status: UNCONFIRMED
          Keywords: missed-optimization
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: pinskia at gcc dot gnu.org
  Target Milestone: ---

Take:
```
void fun (unsigned *x, unsigned *b) {
   asm volatile(""::"r"(*x));
  unsigned *a = b;//malloc (*x);
  if (a == 0)
    return;
  if (a != x)            // (A)
    *a = *x;
  *x = *a;
}
```
GCC currently produces:
  <bb 2> [local count: 1073741824]:
  _1 = *x_6(D);
  __asm__ __volatile__("" :  : "r" _1);
  if (b_7(D) == 0B)
    goto <bb 7>; [18.09%]
  else
    goto <bb 3>; [81.91%]

  <bb 3> [local count: 879501929]:
  if (x_6(D) != b_7(D))
    goto <bb 5>; [70.00%]
  else
    goto <bb 4>; [30.00%]

  <bb 4> [local count: 263850576]:
  pretmp_10 = *b_7(D);
  goto <bb 6>; [100.00%]

  <bb 5> [local count: 615651353]:
  *b_7(D) = _1;

  <bb 6> [local count: 879501929]:
  # prephitmp_11 = PHI <pretmp_10(4), _1(5)>
  *x_6(D) = prephitmp_11;

But pretmp_10 should be the same as _1 as at that point b_7 == x_6.

Note I noticed this while looking into PR 96564.