It's common to represent a $dim$-dimensional vector using a $\lceil \log_2 (dim) \rceil$-qubit register and you get a nice savings on the state preparation cost when you prepare such a state with the rotations + data lookups.
Basically you save on the number of entries you have to look up with the last QROM call so the cost smoothly varies with $dim$, rather than jumping up by a factor of two every time you use another qubit.
I would say yes.
It's common to represent a $
dim
$-dimensional vector using a $\lceil \log_2 (dim) \rceil
$-qubit register and you get a nice savings on the state preparation cost when you prepare such a state with the rotations + data lookups.Basically you save on the number of entries you have to look up with the last QROM call so the cost smoothly varies with $
dim
$, rather than jumping up by a factor of two every time you use another qubit.