Intel's enhanced fused multiply-add operation patent

If an algorithm is non-obvious/inventive enough, I don't see anything wrong with patenting it if patents are valid at all.

But what's really amazing here is that an efficient (or even just any) implementation of the instruction (its interface and contract) that wouldn't be deemed as equivalent to the patented one under the doctrine of equivalents doesn't seem possible. This is an essential patent encumbering the FMA (https://en.wikipedia.org/wiki/Essential_patent).