https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79349
Andrew Pinski <pinskia at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Component|libstdc++ |tree-optimization
--- Comment #5 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
here is the current IR (with -fno-exceptions) at -O3:
MEM[(struct basic_string *)&s] ={v} {CLOBBER};
MEM[(struct _Alloc_hider *)&s] ={v} {CLOBBER};
MEM[(struct _Alloc_hider *)&s]._M_p = &s.D.26433._M_local_buf;
__builtin_memcpy (&s.D.26433._M_local_buf, "abc", 3);
s._M_string_length = 3;
MEM[(char_type &)&s + 19] = 0;
foo ();
_3 = s._M_dataplus._M_p;
if (&s.D.26433._M_local_buf != _3)
goto <bb 3>; [53.47%]
else
goto <bb 4>; [46.53%]
<bb 3> [local count: 574129753]:
_7 = s.D.26433._M_allocated_capacity;
_8 = _7 + 1;
operator delete (_3, _8);
for some reason GCC thinks foo can modify s so the load for s._M_dataplus._M_p
is not be optimized to just &s.D.26433._M_local_buf
