https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125285
Drea Pinski <pinskia at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Ever confirmed|0 |1
Last reconfirmed| |2026-05-12
Status|UNCONFIRMED |NEW
Blocks| |121364
--- Comment #3 from Drea Pinski <pinskia at gcc dot gnu.org> ---
So there is a missed optimization here.
```
MEM <bool Pred::<T427> (const struct Pred *, char) *> [(struct *)&D.3303] =
check;
MEM <long int> [(struct *)&D.3303 + 8B] = 0;
b.f_ = VIEW_CONVERT_EXPR<struct cmf3>(D.3303);
drive_end.0_1 = drive_end;
pred = b;
...
goto <bb 6>; [100.00%]
<bb 3> [local count: 8687547538]:
...
SR.5_10 = MEM[(const struct cmf3 *)&pred].f_.__pfn;
...
<bb 6> [local count: 9761289362]:
if (drive_end.0_1 != 0B)
goto <bb 3>; [89.00%]
else
goto <bb 7>; [11.00%]
```
The load to SR.5_10 should be the same as &check.
pred never escapes (likewise of b).
So there is a missing FRE here.
Now there is also a missing copy prop for aggregate here too:
b.f_ = VIEW_CONVERT_EXPR<struct cmf3>(D.3304);
pred = b;
I will look at it via that angle in later today.
Referenced Bugs:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121364
[Bug 121364] [meta-bug] copy prop for aggregates
