Struct BlockQ8_0 

#[repr(C)]
pub struct BlockQ8_0 {
    pub d: f16,
    pub qs: [i8; 32],
}

Q8_0 block: 32 weights quantized to 8-bit signed integers with a shared FP16 scale.

Layout (34 bytes):

• d: FP16 block scale.
• qs: 32 bytes of i8 quantized weights.

Dequant: w[j] = d × qs[j] as f32.

Fields

d: f16

Block scale (FP16).

qs: [i8; 32]

32 × int8 quantized weights.

Implementations

impl BlockQ8_0

pub fn dequant(blocks: &[Self], output: &mut [f32]) -> BonsaiResult<()>

Dequantize a slice of Q8_0 blocks into f32 output.

output must have length >= blocks.len() * QK_Q8_0.

pub fn quantize(input: &[f32]) -> BonsaiResult<Vec<Self>>

Quantize f32 input into Q8_0 blocks.

Input length must be a multiple of QK_Q8_0 (32).

Scale = max(|x|) / 127; qs[j] = clamp(round(x / scale), -127, 127).

pub fn slice_from_bytes(data: &[u8]) -> BonsaiResult<&[Self]>

Zero-copy cast of a byte slice to a slice of BlockQ8_0.

Errors

Returns BonsaiError::KQuantError if data.len() is not a multiple of BLOCK_Q8_0_BYTES or the pointer is not 2-byte aligned (required for the embedded f16 field).

pub fn dequant_to_buf(&self, buf: &mut [f32; 32])

Dequantize this single block into a 32-element f32 buffer.

Used by the GEMV kernel to avoid heap allocation on the hot path.

Trait Implementations

impl Clone for BlockQ8_0

fn clone(&self) -> BlockQ8_0

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for BlockQ8_0

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

Formats the value using the given formatter.

impl PartialEq for BlockQ8_0

fn eq(&self, other: &BlockQ8_0) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Copy for BlockQ8_0

impl StructuralPartialEq for BlockQ8_0