https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102569
Bug ID: 102569
Summary: Missed redudant add with add_overflow on the tree
level
Product: gcc
Version: 12.0
Status: UNCONFIRMED
Keywords: missed-optimization, TREE
Severity: enhancement
Priority: P3
Component: tree-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: pinskia at gcc dot gnu.org
Target Milestone: ---
Take:
bool f1(unsigned x, unsigned y, unsigned *res)
{
bool t = __builtin_add_overflow(x, y, res);
*res -= (x+y);
return t;
}
----- CUT ----
We currently get:
_7 = .ADD_OVERFLOW (x_5(D), y_6(D));
_1 = REALPART_EXPR <_7>;
_2 = IMAGPART_EXPR <_7>;
t_10 = (bool) _2;
_3 = x_5(D) + y_6(D);
_4 = _1 - _3;
But _3 and _1 are the same _4 should be optimized to 0 as it is on the RTL
level.
The way I see this being done is during the late fre add x+y in the table for
REALPART_EXPR<_7> and it will be done correctly.
