Struct PQDistanceTable 

pub struct PQDistanceTable<S: EmbeddingSpace, const M: usize, const NBITS: usize>
where
    [(); { _ }]:,
{ /* private fields */ }

Precomputed distance table for Asymmetric Distance Computation (ADC).

Given a query vector, this table stores the squared distance from each query subvector to each centroid in that subspace. During search, the distance to an encoded vector is computed by summing table lookups.

Table layout: table[m * ksub + k] is the distance from query subvector m to centroid k in subspace m.

The const generics must match the ProductQuantizer configuration:

• M: number of subquantizers
• NBITS: bits per centroid index

Implementations

impl<S: EmbeddingSpace, const M: usize, const NBITS: usize> PQDistanceTable<S, M, NBITS>

where [(); { _ }]:,

pub const KSUB: usize

Number of centroids per subspace (2^NBITS)

pub fn new(table: Vec<S::DistanceValue>, ksub: usize) -> Self

pub fn distance(&self, code: &PQCode<M, NBITS>) -> S::DistanceValue

Compute approximate distance to an encoded vector using table lookups.

This is the core of ADC: instead of computing the full distance, we sum precomputed partial distances from each subspace.

pub fn m(&self) -> usize

pub fn ksub(&self) -> usize

Trait Implementations

impl<S: EmbeddingSpace, const M: usize, const NBITS: usize> Debug for PQDistanceTable<S, M, NBITS>

where [(); { _ }]:,

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.