https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114589
--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #1)
> Confirmed.
>
> Why didn't sink1 push _10, _13, _12, and _11 past the conditional here ...
> If it did that I think it might have optimized correctly.
Because then _12 would be 1, _13 would be 4 and _10 would be come `_11 + 4` and
the loop induction variable would be going from _11 to `_11 + 4` then:
```
# __for_begin_15 = PHI <__for_begin_8(6), _11(5)>
# .MEM_16 = PHI <.MEM_7(6), .MEM_4(D)(5)>
# VUSE <.MEM_16>
i_6 = *__for_begin_15;
# .MEM_7 = VDEF <.MEM_16>
# USE = nonlocal escaped
# CLB = nonlocal escaped
_Z1fiD.2782 (i_6);
# PT = nonlocal
__for_begin_8 = __for_begin_15 + 4;
if (__for_begin_8 != _10)
goto <bb 6>; [89.00%]
else
goto <bb 10>; [11.00%]
```
And we could figure out this loop is only gone through once only.
