To the trolls that downvoted my post without explaining why, here is the proof: https://godbolt.org/g/xXecW2

I rewrote the function using if statements making it compile to exactly the same instructions.

If you optimize the code with compiler flags (e.g: -msse, -msse2 for SSE/SSE2, -O3 for optimization) the generated instructions are exactly the same.

My point being: the code in the blog post is not branchless. The absence of if statements doesn't mean it's branchless.

If you care to disagree here by downvoting I dare you to reply explaining why you think this is wrong.