Struct CpuSimd 

pub struct CpuSimd<const WIDTH: usize>;

Portable CPU SIMD platform

This uses scalar emulation for portability. In a real implementation, this would use std::simd or platform-specific intrinsics (AVX-512, NEON).

Trait Implementations

impl<const WIDTH: usize> Clone for CpuSimd<WIDTH>

fn clone(&self) -> CpuSimd<WIDTH>

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl<const WIDTH: usize> Debug for CpuSimd<WIDTH>

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

Formats the value using the given formatter.

impl<const WIDTH: usize> Platform for CpuSimd<WIDTH>

where [(); WIDTH]: Sized,

const WIDTH: usize = WIDTH

Number of parallel lanes

const NAME: &'static str = "CpuSimd"

Platform name for debugging

type Vector<T: GpuValue> = PortableVector<T, WIDTH>

The vector type for this platform

type Mask = u64

Mask type for predicated operations

fn shuffle<T: GpuValue>( source: Self::Vector<T>, indices: Self::Vector<u32>, ) -> Self::Vector<T>

Shuffle: each lane reads from source[indices[lane]]

fn shuffle_down<T: GpuValue>( source: Self::Vector<T>, delta: usize, ) -> Self::Vector<T>

Shuffle down: lane i reads from lane i+delta (with wrapping or clamping)

fn shuffle_xor<T: GpuValue>( source: Self::Vector<T>, mask: usize, ) -> Self::Vector<T>

Shuffle XOR: lane i reads from lane i^mask

fn reduce_sum<T: GpuValue + Add<Output = T>>(values: Self::Vector<T>) -> T

Sum all lanes, result available in all lanes (uniform)

fn reduce_max<T: GpuValue + Ord>(values: Self::Vector<T>) -> T

Maximum across all lanes

fn reduce_min<T: GpuValue + Ord>(values: Self::Vector<T>) -> T

Minimum across all lanes

fn ballot(predicates: Self::Vector<bool>) -> Self::Mask

Ballot: collect per-lane bools into a mask

fn all(predicates: Self::Vector<bool>) -> bool

All lanes true?

fn any(predicates: Self::Vector<bool>) -> bool

Any lane true?

fn mask_popcount(mask: Self::Mask) -> u32

Population count of a mask

fn broadcast<T: GpuValue>(value: T) -> Self::Vector<T>

Broadcast a scalar to all lanes

impl<const WIDTH: usize> Copy for CpuSimd<WIDTH>