Struct CpuAffinityInput 

pub struct CpuAffinityInput {
    pub domain: Option<CpuDomainId>,
    pub logical_bytes: usize,
}

CPU affinity metadata for one logical operation input.

The resolver reads this metadata only. It never changes, copies, or rehomes tensor payloads.

Examples

use tenferro_cpu::CpuAffinityInput;
use tenferro_tensor::CpuDomainId;

let input = CpuAffinityInput {
    domain: Some(CpuDomainId::new(3)),
    logical_bytes: 64,
};
assert_eq!(input.logical_bytes, 64);

Fields

domain: Option<CpuDomainId>

Known CPU execution domain, or None when affinity is unknown.

logical_bytes: usize

Logical input size used by CpuAffinityPolicy::DominantInputBytes.

Implementations

impl CpuAffinityInput

pub fn from_tensor(tensor: &Tensor) -> Result<Self, CpuAffinityInputError>

Construct resolver input metadata from a tensor.

The logical byte count is the checked shape product times the tensor’s scalar byte width. CPU affinity is copied from placement metadata; the tensor and its storage are otherwise untouched.

Examples

use tenferro_cpu::CpuAffinityInput;
use tenferro_tensor::Tensor;

let tensor = Tensor::from_vec_col_major(vec![2], vec![1.0_f64, 2.0])?;
let input = CpuAffinityInput::from_tensor(&tensor)?;
assert_eq!(input.logical_bytes, 2 * std::mem::size_of::<f64>());

Errors

Returns CpuAffinityInputError when the logical element or byte count cannot be represented by usize.

pub fn from_parts( domain: Option<CpuDomainId>, shape: &[usize], dtype: DType, ) -> Result<Self, CpuAffinityInputError>

Construct resolver input metadata from placement, shape, and dtype.

Scalar shapes have one element. Any zero extent yields zero logical bytes without multiplying the other extents.

Examples

use tenferro_cpu::CpuAffinityInput;
use tenferro_tensor::DType;

let scalar = CpuAffinityInput::from_parts(None, &[], DType::F32)?;
let empty = CpuAffinityInput::from_parts(None, &[usize::MAX, 0], DType::F64)?;
assert_eq!(scalar.logical_bytes, 4);
assert_eq!(empty.logical_bytes, 0);

Errors

Returns CpuAffinityInputError::ShapeProductOverflow when non-zero extents overflow, or CpuAffinityInputError::LogicalByteCountOverflow when multiplying by the dtype width overflows.

Trait Implementations

impl Clone for CpuAffinityInput

fn clone(&self) -> CpuAffinityInput

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for CpuAffinityInput

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

Formats the value using the given formatter.

impl PartialEq for CpuAffinityInput

fn eq(&self, other: &CpuAffinityInput) -> 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 Copy for CpuAffinityInput

impl Eq for CpuAffinityInput

impl StructuralPartialEq for CpuAffinityInput