[Bug tree-optimization/114040] wrong code with __builtin_sub_overflow_p() and _BitInt() at -O2 and above

[jakub at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114040

--- Comment #6 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Ah, I actually see the what's wrong.
The destination store for the first limb in the loop look correct:
  if (_14 <= 3)
    goto <bb 6>; [80.00%]
  else
    goto <bb 7>; [20.00%]

  <bb 6> [local count: 1073741824]:
  bitint.3[_14] = _21;

  <bb 7> [local count: 1073741824]:
i.e. if _14 is 0 or 2, we store something, otherwise we don't, because result
__real__ only has 4 limbs.
But the second limb store is wrong:
  if (_27 <= 3)
    goto <bb 12>; [80.00%]
  else
    goto <bb 15>; [20.00%]

  <bb 12> [local count: 1073741824]:
  if (_27 < 3)
    goto <bb 14>; [80.00%]
  else
    goto <bb 13>; [20.00%]

  <bb 13> [local count: 1073741824]:
  bitint.3[_27] = _30;
  goto <bb 15>; [100.00%]

  <bb 14> [local count: 858993464]:
  MEM[(unsigned long *)&bitint.3 + 24B] = _30;

  <bb 15> [local count: 1073741824]:
It should be the other way around, if _27 < 3 (aka if it is 1), store
bitint.3[_27] = _30; , if it is 3, MEM[(unsigned long *)&bitint.3 + 24B] = _30;
and if it is > 3, don't store anything.  I'm also surprised it doesn't mask off
the most significant bit.