> Optimization is not only rearranging instructions.

And not only finding parallelizable instructions either, I assume.

Thanks for the clarification, because such a proof would have surprised me a lot.