I can’t find any assertion in the post about the content of any of the input data sets or search keys, other than that they’re “unpredictable”. I assume purely

I can’t find any assertion in the post about the content of any of the input data sets or search keys, other than that they’re “unpredictable”. I assume purely random, but if these 8-byte strings were anything but pure information, modern branch predictors can sniff their way to better performance than cmov pretty easily.