pub fn encode_turbovec(
vector: &[f32],
bits: u8,
) -> Result<EncodedVector, EncodingError>Expand description
Encode a single f32 vector under the turbovec codec at
bits bit-width.
The returned EncodedVector holds the source vector’s
f32 little-endian bytes; the codec marker on the
EncodedVector indicates that the table-level SIMD index
owns the compressed representation. See module docs for the
rationale.
§Errors
EncodingError::UnsupportedBitWidth for bits outside
{2, 3, 4}; EncodingError::EmptyVector for a zero-dim
input; EncodingError::UnsupportedDim when dim is not a
positive multiple of 8; EncodingError::DimensionTooLarge
for >65535 dimensions; EncodingError::NonFinite for any
non-finite component.