[Bug tree-optimization/82282] PRE cannot blindly fold integer-to-pointer/pointer-to-integer round-trips

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

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82282

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Depends on|                            |82177

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
This is really the same / similar as PR82177.

  if (u_14 != v_15)
    goto <bb 3>;
  else
    goto <bb 4>;

  <bb 4>:
  # v_1 = PHI <u_14(2), u_14(3)>
  v.0_19 = (int *) v_1;
  glb = v.0_19;

so there's a missed trivial optimization not done by phiopt to

  u_14 = (uintptr_t) &MEM[(void *)&x + 4B];
...
  if (u_14 != v_15)
    goto <bb 3>;
  else
    goto <bb 4>;

  <bb 4>:
  v.0_19 = (int *) u_14;

where (int *) u_14 would be folded to just &MEM[(void *)&x + 4B].

It's not partial redundancy elimination but value-numbering does this
kind of cleanup.  forwprop would also do it and there's nothing wrong
with this.

In C you are only allowed to convert a pointer back and forth
with [u]intptr_t, so when you convert it back you get the original
pointer which means it still points to one-after 'x'.  You can't
make it possibly point to 'y' by this trick.  Specifically you
are not allowed to make a uintptr_t from a pointer to object A,
do some magic on that value, convert it to a pointer again and
expect it to point to object B.


Referenced Bugs:

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82177
[Bug 82177] Alias analysis too aggressive with integer-to-pointer cast