https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125693
Drea Pinski <pinskia at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Blocks| |121364
Status|UNCONFIRMED |NEW
Severity|normal |enhancement
Keywords| |missed-optimization
Ever confirmed|0 |1
Component|libstdc++ |tree-optimization
Last reconfirmed| |2026-06-10
--- Comment #3 from Drea Pinski <pinskia at gcc dot gnu.org> ---
So at -O2 we could optimize this (I have not looked into libstdc++ sources).
We have
# .MEM_20 = VDEF <.MEM_19>
MEM[(struct array *)&__tmp + 8B] = {};
...
# .MEM_78 = PHI <.MEM_86(6), .MEM_85(7), .MEM_87(8)>
# .MEM_88 = VDEF <.MEM_78>
MEM[(struct array *)__position_1] = MEM[(struct array &)&__tmp + 8];
But currently the optimization for zero prop stops at phi nodes.
So this is missed out. I have some idea on how to fix this but currently not
working on it.
clang is able to handle this at -O1+.
Note with libc++, LLVM fails at all optimizations level.
So confirmed for the missing gimple level optimization.
Referenced Bugs:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121364
[Bug 121364] [meta-bug] copy prop for aggregates
