Struct TensorLayout 

pub struct TensorLayout {
    pub strides: Option<Vec<i64>>,
    pub format: MemoryFormat,
    pub alignment: usize,
}

First-class strided layout for a value.

strides == None means “contiguous row-major for the value’s shape”; this is the common case and avoids materializing strides for every value.

Fields

strides: Option<Vec<i64>>

Physical strides in elements. None == contiguous row-major.

format: MemoryFormat

Memory-format hint.

alignment: usize

Required alignment in bytes for the backing allocation.

Implementations

impl TensorLayout

pub fn contiguous() -> Self

A contiguous row-major layout (strides implied by shape).

pub fn strided(strides: Vec<i64>) -> Self

A layout with explicit strides (marked MemoryFormat::Custom).

pub fn is_contiguous(&self, shape: &[usize]) -> bool

Whether this layout is contiguous row-major for shape.

pub fn resolved_strides(&self, shape: &[usize]) -> Vec<i64>

The strides for shape under this layout, materializing the implied contiguous strides when strides == None.

pub fn transpose(&self, shape: &[usize], perm: &[usize]) -> Self

Reorder axes without copying data (a lazy transpose).

pub fn storage_size(&self, shape: &[usize], dtype: DataType) -> usize

Total backing storage size in bytes: the largest byte offset reachable via the strides, plus one element. Handles negative strides.

Trait Implementations

impl Clone for TensorLayout

fn clone(&self) -> TensorLayout

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for TensorLayout

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

Formats the value using the given formatter.

impl Default for TensorLayout

fn default() -> Self

Returns the “default value” for a type.

impl PartialEq for TensorLayout

fn eq(&self, other: &TensorLayout) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for TensorLayout