Missing the mask shift from FastCDC paper?

[jokkebk]

Nice compact implementation. However, I fail to see the "spacing" of the FastCDC paper implemented, which I think degrades the quality (as explained in the paper).

So instead of a mask like this (4 bits high):

maskS = 0b100100010001000

You create this:

maskS = 0b000000000001111

Some additional math to spread out the one bits would be needed.