When I see a snippet like this int temp = a * b; ngassert( abs(temp) >= abs(a), "overflow!" ); I wonder what stops the compiler from opt

When I see a snippet like this int temp = a * b; ngassert( abs(temp) >= abs(a), "overflow!" ); I wonder what stops the compiler from optimizing away the assert, as it's clearly based on an undefined behavior.