Struct SharedTile 

pub struct SharedTile<T, const W: usize, const H: usize> { /* private fields */ }

A 2D shared memory tile.

This type represents a 2D array in CUDA shared memory. On the CPU side, it’s a zero-sized type that serves as a marker for the transpiler.

Type Parameters

• T - Element type (f32, f64, i32, etc.)
• W - Tile width (columns)
• H - Tile height (rows)

Example

// 16x16 tile of floats
let tile = SharedTile::<f32, 16, 16>::new();

// 32x8 tile for matrix operations
let mat_tile = SharedTile::<f32, 32, 8>::new();

Implementations

impl<T: Default + Copy, const W: usize, const H: usize> SharedTile<T, W, H>

pub fn new() -> Self

Create a new shared memory tile.

On GPU, this translates to: __shared__ T tile[H][W];

pub const fn width() -> usize

Get the tile width.

pub const fn height() -> usize

Get the tile height.

pub const fn size() -> usize

Get the total number of elements.

pub fn get(&self, _x: i32, _y: i32) -> T

Get an element from the tile (CPU stub - actual access on GPU).

On GPU, this translates to: tile[y][x]

Arguments

• x - Column index (0..W)
• y - Row index (0..H)

pub fn set(&mut self, _x: i32, _y: i32, _value: T)

Set an element in the tile (CPU stub - actual write on GPU).

On GPU, this translates to: tile[y][x] = value;

Arguments

• x - Column index (0..W)
• y - Row index (0..H)
• value - Value to store

Trait Implementations

impl<T: Debug, const W: usize, const H: usize> Debug for SharedTile<T, W, H>

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

Formats the value using the given formatter.

impl<T: Default + Copy, const W: usize, const H: usize> Default for SharedTile<T, W, H>

fn default() -> Self

Returns the “default value” for a type.