Why does Clang optimize away x * 1.0 but NOT x + 0.0?

Question

Why does Clang optimize away the loop in this code

#include 
#include 

static size_t const N = 1 << 27;
static double arr         


        

Answer 1

x += 0.0 isn't a NOOP if x is -0.0. The optimizer could strip out the whole loop anyway since the results aren't used, though. In general, it's hard to tell why an optimizer makes the decisions it does.