It could be a very wide area we're actually searching in the index (if the bitvector is very sparse of whichever kind of bit we are selecting on) and we can probably tell quite easily how sparse by looking at the next sample if there is one (or the end of the L0 block) to get an upper bound for the binary search.
It could be a very wide area we're actually searching in the index (if the bitvector is very sparse of whichever kind of bit we are selecting on) and we can probably tell quite easily how sparse by looking at the next sample if there is one (or the end of the L0 block) to get an upper bound for the binary search.