Struct TiledReducer 

pub struct TiledReducer {
    pub tile_size: usize,
}

Parallel tiled reducer.

Splits the input into tiles of tile_size, performs a local reduction on each tile in parallel, then reduces the per-tile results serially.

Fields

tile_size: usize

Number of elements per tile (analogous to GPU work-group size).

Implementations

impl TiledReducer

pub fn new(tile_size: usize) -> Self

Create a new tiled reducer.

pub fn sum(&self, data: &[f64]) -> f64

Compute the global sum using two-level tiled reduction.

pub fn max(&self, data: &[f64]) -> f64

Compute the global maximum.

pub fn min(&self, data: &[f64]) -> f64

Compute the global minimum.

pub fn dot(&self, a: &[f64], b: &[f64]) -> f64

Compute the global dot product of two equal-length slices.

pub fn tile_sums(&self, data: &[f64]) -> Vec<f64>

Compute per-tile sums (exposes intermediate tile results).

Trait Implementations

impl Clone for TiledReducer

fn clone(&self) -> TiledReducer

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for TiledReducer

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

Formats the value using the given formatter.