https://gcc.gnu.org/bugzilla/show_bug.cgi?id=123067
Andrew Pinski <pinskia at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Summary|wrong code at -O2 and above |[16 Regression] wrong code
|on x86_64-pc-linux-gnu |at -O2 and above on
| |x86_64-pc-linux-gnu
Target Milestone|--- |16.0
Status|UNCONFIRMED |NEW
Last reconfirmed| |2025-12-09
Ever confirmed|0 |1
--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Confirmed.
ldist turns it into a "memmove":
```
<bb 3> [local count: 118111600]:
# i.6_23 = PHI <_6(7), 1(2)>
# ivtmp_18 = PHI <ivtmp_19(7), 255(2)>
_7 = &a + 4;
_5 = &a + 8;
_27 = MEM <uint128_t> [(char * {ref-all})_5];
MEM <uint128_t> [(char * {ref-all})_7] = _27;
_6 = i.6_23 + 1;
ivtmp_19 = ivtmp_18 - 1;
if (ivtmp_19 != 0)
goto <bb 7>; [89.00%]
else
goto <bb 6>; [11.00%]
<bb 7> [local count: 105119324]:
goto <bb 3>; [100.00%]
```
Which is fine. But then LIM4 decides move the load out of the loop which is
wrong:
Moving statement _27 = MEM <uint128_t> [(char * {ref-all})&a + 8B];
(cost 20) out of loop 1.
