Struct Fp16Codec 

pub struct Fp16Codec { /* private fields */ }

IEEE 754 binary16: 2 bytes/dim. Hot-tier codec (~1e-3 relative error). Conversion is implemented in-crate (no half dependency) with round-to-nearest-even, matching hardware semantics.

Implementations

impl Fp16Codec

pub fn new(dim: usize, metric: DistanceMetric) -> Self

Trait Implementations

impl VectorCodec for Fp16Codec

fn kind(&self) -> CodecKind

Which representation this codec produces.

fn dim(&self) -> usize

Dimensionality of vectors this codec accepts.

fn encoded_len(&self) -> usize

Bytes per encoded vector.

fn encode(&self, v: &[f32]) -> Result<Vec<u8>>

Encode a vector. The blob is self-sufficient for scoring.

fn decode(&self, blob: &[u8]) -> Result<Vec<f32>>

Reconstruct an approximation (exact only for Fp32).

fn distance(&self, a: &[u8], b: &[u8]) -> Result<f32>

Symmetric blob×blob distance under this codec’s metric.

fn make_query(&self, q: &[f32]) -> Result<Box<dyn EncodedQuery>>

Prepare a query for repeated asymmetric scoring against blobs.