Struct SimdQuantizationOps 

pub struct SimdQuantizationOps { /* private fields */ }

SIMD-accelerated quantization operations

This struct provides vectorized implementations of quantization operations that can take advantage of CPU SIMD instructions for improved performance.

Implementations

impl SimdQuantizationOps

pub fn new() -> Self

Create new SIMD quantization operations

pub fn quantize_f32_to_u8_simd( &self, input: &[f32], scale: f32, zero_point: f32, ) -> BackendResult<Vec<u8>>

SIMD-accelerated f32 to u8 quantization

Uses vectorized operations to quantize multiple floating-point values to 8-bit unsigned integers simultaneously.

pub fn dequantize_u8_to_f32_simd( &self, input: &[u8], scale: f32, zero_point: f32, ) -> BackendResult<Vec<f32>>

SIMD-accelerated u8 to f32 dequantization

pub fn add_int8_simd(&self, a: &[i8], b: &[i8]) -> BackendResult<Vec<i8>>

SIMD-accelerated INT8 vector addition

pub fn is_available(&self) -> bool

Check if SIMD operations are available

pub fn vector_width(&self) -> usize

Get the optimal vector width for this hardware

Trait Implementations

impl Clone for SimdQuantizationOps

fn clone(&self) -> SimdQuantizationOps

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for SimdQuantizationOps

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

Formats the value using the given formatter.