https://gcc.gnu.org/bugzilla/show_bug.cgi?id=127970
Bug ID: 127970
Summary: wrong code at -O1 and above: _Decimal64 (x == C) ? C :
x returns x, changing the quantum exponent
Product: gcc
Version: 17.0
Status: UNCONFIRMED
Keywords: wrong-code
Severity: normal
Priority: P3
Component: tree-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: kyuwoncho18 at gmail dot com
Target Milestone: ---
Target: x86_64-pc-linux-gnu
int main(void) {
volatile _Decimal64 v = 1.00DD;
_Decimal64 x = v, c = 1.0DD, r = x == c ? c : x;
return __builtin_memcmp(&r, &c, sizeof r) != 0;
}
godbolt: https://godbolt.org/z/rzPE41qcT
With -O0 it returns 0, but with GCC -O{1,2,3,s,g} it returns 1: x == 1.0DD is
true, but r gets x (1.00DD, equal value, different quantum exponent) instead of
the constant 1.0DD. -fno-tree-dominator-opts avoids it.
Disclaimer: This test case was generated by a fuzzer, and the report was
written with LLM assistance.
