https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98703
Bug ID: 98703
Summary: Failure to optimize out non-zero check relative to
multiplication overflow check
Product: gcc
Version: 11.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: tree-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: gabravier at gmail dot com
Target Milestone: ---
bool f1(unsigned x, unsigned y, unsigned *res)
{
return __builtin_mul_overflow(x, y, res) && x;
}
This can be optimized to `return __builtin_mul_overflow(x, y, res);`. This
transformation is done by LLVM, but not by GCC.
PS: I originally found this while looking at the code generation for this code
(from https://gcc.gnu.org/PR95852) :
bool f2(unsigned x, unsigned y, unsigned *res)
{
*res = x * y;
return x && ((*res / x) != y);
}
f2 is equivalent to `return __builtin_mul_overflow(x, y, res);`, but the code
emitted is more like f1.
