mrdanielps Properly implement (most) floating-point instructions