Struct CpuPure 

pub struct CpuPure;

Trait Implementations

impl Backend for CpuPure

const SUPPORTS_2D_TRANSPOSED_MATMUL: bool = true

true if OpKind::MatMul accepts a 2D input whose strides describe a transposed view (row_stride == 1, col_stride > 1) - i.e. the underlying GEMM is invoked with a trans-flag and no copy is required. The fast path is deliberately scoped to rank 2; higher-rank tensors are always contiguified by the op layer before reaching the kernel.

const SUPPORTS_NON_CONTIGUOUS_MATMUL: bool = true

true if OpKind::MatMul accepts any memory configuration as long the last 2 axis are contiguous.

fn compute<T>( op: &OpKind<T>, output_buffer: Vec<T>, output_layout: &Layout, inputs: &[TensorData<T>], ) -> TensorData<T>

where T: Dtype + ComputeFor<CpuPure>,

Run op over inputs into a fresh allocation. output_buffer is the destination Vec<T>; the returned TensorData wraps it with output_layout.

fn compute_inplace<T>( op: &OpKind<T>, output_layout: &Layout, inputs: Vec<TensorData<T>>, output_idx: usize, ) -> TensorData<T>

where T: Dtype + ComputeFor<Self>,

Run op reusing inputs[output_idx]’s buffer as the destination. The planner guarantees that buffer is no longer referenced by any live node at this point, so the in-place write is sound.

impl Clone for CpuPure

fn clone(&self) -> CpuPure

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl ComputeFor<CpuPure> for f64

fn compute( op: &OpKind<f64>, output_buffer: Vec<f64>, output_layout: &Layout, inputs: &[TensorData<f64>], ) -> TensorData<f64>

fn compute_inplace( op: &OpKind<Self>, output_layout: &Layout, inputs: Vec<TensorData<Self>>, output_idx: usize, ) -> TensorData<Self>

impl ComputeFor<CpuPure> for f32

fn compute( op: &OpKind<f32>, output_buffer: Vec<f32>, output_layout: &Layout, inputs: &[TensorData<f32>], ) -> TensorData<f32>

fn compute_inplace( op: &OpKind<Self>, output_layout: &Layout, inputs: Vec<TensorData<Self>>, output_idx: usize, ) -> TensorData<Self>

impl Copy for CpuPure

impl Debug for CpuPure

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

Formats the value using the given formatter.

impl Default for CpuPure

fn default() -> CpuPure

Returns the “default value” for a type.

impl Eq for CpuPure

impl Hash for CpuPure

fn hash<__H: Hasher>(&self, state: &mut __H)

Feeds this value into the given Hasher.

fn hash_slice<H>(data: &[Self], state: &mut H)

where H: Hasher, Self: Sized,

Feeds a slice of this type into the given Hasher.

impl PartialEq for CpuPure

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

Tests for self and other values to be equal, and is used by ==.

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

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl StructuralPartialEq for CpuPure