pub const fn bit_scale<const NBITS: usize>() -> f32
Return the scaling parameter a that should be multiplied to compressed vector codes.
a