It's worth noting that...

AlignRight is exposed as (left, right, mask) where left is the upper 128-bits and right is the lower 128-bits of the concatenated vector, therefore mask 1 picks right[1]-right[15] and left[0] as the last element.

AdvSimd.ExtractVector128 is exposed as (upper, lower, mask), but it should've been (lower, upper, mask), so its inverse from the AlignRight order. This can be potentially confusing when reviewing the code.

The PR here is doing it correctly, just wanted to call out that its a weird thing if you go look at stuff