Comment by bnegreve

I don't see how generating different code for the same mathematical expression can be a good thing.

The compiler should detect that the two expressions are strictly equivalent and generate whatever code it believes is the fastest.

Any idea why it is this way?