Trait SimdOps 

pub trait SimdOps: Send + Sync {
    // Required methods
    fn name(&self) -> &'static str;
    fn is_available(&self) -> bool;
    fn add_i16x8(&self, a: I16x8, b: I16x8) -> I16x8;
    fn sub_i16x8(&self, a: I16x8, b: I16x8) -> I16x8;
    fn mul_i16x8(&self, a: I16x8, b: I16x8) -> I16x8;
    fn add_i32x4(&self, a: I32x4, b: I32x4) -> I32x4;
    fn sub_i32x4(&self, a: I32x4, b: I32x4) -> I32x4;
    fn min_i16x8(&self, a: I16x8, b: I16x8) -> I16x8;
    fn max_i16x8(&self, a: I16x8, b: I16x8) -> I16x8;
    fn clamp_i16x8(&self, v: I16x8, min: i16, max: i16) -> I16x8;
    fn min_u8x16(&self, a: U8x16, b: U8x16) -> U8x16;
    fn max_u8x16(&self, a: U8x16, b: U8x16) -> U8x16;
    fn clamp_u8x16(&self, v: U8x16, min: u8, max: u8) -> U8x16;
    fn horizontal_sum_i16x8(&self, v: I16x8) -> i32;
    fn horizontal_sum_i32x4(&self, v: I32x4) -> i32;
    fn sad_u8x16(&self, a: U8x16, b: U8x16) -> u32;
    fn sad_8(&self, a: &[u8], b: &[u8]) -> u32;
    fn sad_16(&self, a: &[u8], b: &[u8]) -> u32;
    fn widen_low_u8_to_i16(&self, v: U8x16) -> I16x8;
    fn widen_high_u8_to_i16(&self, v: U8x16) -> I16x8;
    fn narrow_i32x4_to_i16x8(&self, low: I32x4, high: I32x4) -> I16x8;
    fn madd_i16x8(&self, a: I16x8, b: I16x8, c: I16x8) -> I16x8;
    fn pmaddwd(&self, a: I16x8, b: I16x8) -> I32x4;
    fn shr_i16x8(&self, v: I16x8, shift: u32) -> I16x8;
    fn shl_i16x8(&self, v: I16x8, shift: u32) -> I16x8;
    fn shr_i32x4(&self, v: I32x4, shift: u32) -> I32x4;
    fn shl_i32x4(&self, v: I32x4, shift: u32) -> I32x4;
    fn avg_u8x16(&self, a: U8x16, b: U8x16) -> U8x16;
}

Core SIMD operations trait.

This trait defines the fundamental SIMD operations needed for video codec implementations. All operations are designed to map efficiently to hardware SIMD instructions.

Required Methods

fn name(&self) -> &'static str

Get the name of this SIMD implementation.

fn is_available(&self) -> bool

Check if this implementation is available on the current CPU.

fn add_i16x8(&self, a: I16x8, b: I16x8) -> I16x8

Element-wise addition of two i16x8 vectors.

fn sub_i16x8(&self, a: I16x8, b: I16x8) -> I16x8

Element-wise subtraction of two i16x8 vectors.

fn mul_i16x8(&self, a: I16x8, b: I16x8) -> I16x8

Element-wise multiplication of two i16x8 vectors.

fn add_i32x4(&self, a: I32x4, b: I32x4) -> I32x4

Element-wise addition of two i32x4 vectors.

fn sub_i32x4(&self, a: I32x4, b: I32x4) -> I32x4

Element-wise subtraction of two i32x4 vectors.

fn min_i16x8(&self, a: I16x8, b: I16x8) -> I16x8

Element-wise minimum of two i16x8 vectors.

fn max_i16x8(&self, a: I16x8, b: I16x8) -> I16x8

Element-wise maximum of two i16x8 vectors.

fn clamp_i16x8(&self, v: I16x8, min: i16, max: i16) -> I16x8

Element-wise clamp of i16x8 vector.

fn min_u8x16(&self, a: U8x16, b: U8x16) -> U8x16

Element-wise minimum of two u8x16 vectors.

fn max_u8x16(&self, a: U8x16, b: U8x16) -> U8x16

Element-wise maximum of two u8x16 vectors.

fn clamp_u8x16(&self, v: U8x16, min: u8, max: u8) -> U8x16

Element-wise clamp of u8x16 vector.

fn horizontal_sum_i16x8(&self, v: I16x8) -> i32

Horizontal sum of all elements in an i16x8 vector.

fn horizontal_sum_i32x4(&self, v: I32x4) -> i32

Horizontal sum of all elements in an i32x4 vector.

fn sad_u8x16(&self, a: U8x16, b: U8x16) -> u32

Sum of absolute differences between two u8x16 vectors.

Computes: sum(|a[i] - b[i]|) for all i

fn sad_8(&self, a: &[u8], b: &[u8]) -> u32

Sum of absolute differences for 8 bytes.

fn sad_16(&self, a: &[u8], b: &[u8]) -> u32

Sum of absolute differences for 16 bytes.

fn widen_low_u8_to_i16(&self, v: U8x16) -> I16x8

Widen u8x16 low half to i16x8.

fn widen_high_u8_to_i16(&self, v: U8x16) -> I16x8

Widen u8x16 high half to i16x8.

fn narrow_i32x4_to_i16x8(&self, low: I32x4, high: I32x4) -> I16x8

Narrow two i32x4 to i16x8 with saturation.

fn madd_i16x8(&self, a: I16x8, b: I16x8, c: I16x8) -> I16x8

Multiply and add: a * b + c for i16x8.

fn pmaddwd(&self, a: I16x8, b: I16x8) -> I32x4

Multiply pairs and add adjacent results (pmaddwd equivalent).

Multiplies pairs of i16 elements and adds adjacent products: result[0] = a[0]*b[0] + a[1]*b[1] result[1] = a[2]*b[2] + a[3]*b[3] etc.

fn shr_i16x8(&self, v: I16x8, shift: u32) -> I16x8

Arithmetic right shift of i16x8 by immediate.

fn shl_i16x8(&self, v: I16x8, shift: u32) -> I16x8

Logical left shift of i16x8 by immediate.

fn shr_i32x4(&self, v: I32x4, shift: u32) -> I32x4

Arithmetic right shift of i32x4 by immediate.

fn shl_i32x4(&self, v: I32x4, shift: u32) -> I32x4

Logical left shift of i32x4 by immediate.

fn avg_u8x16(&self, a: U8x16, b: U8x16) -> U8x16

Average of two u8x16 vectors (rounding up).

Implementors

impl SimdOps for NeonSimd

impl SimdOps for ScalarFallback

impl SimdOps for Avx2Simd

impl SimdOps for Avx512Simd