Struct TensorRef 

#[non_exhaustive]
pub struct TensorRef {
    pub name: Ident,
    pub dtype: DataType,
    pub shape: Arc<[u32]>,
}

A named, typed, shaped buffer argument passed into a Cat-A op.

Construct with TensorRef::new or the convenience helpers (u32_1d, f32_1d, u32_2d, f32_2d). Downstream ops consume TensorRefs instead of raw &str buffer names so type + shape checks happen at build() time.

Fields (Non-exhaustive)

Non-exhaustive structs could have additional fields added in future. Therefore, non-exhaustive structs cannot be constructed in external crates using the traditional Struct { .. } syntax; cannot be matched against without a wildcard ..; and struct update syntax will not work.

name: Ident

Name the buffer is registered under. Matches BufferDecl::name.

dtype: DataType

Element dtype. Enforced against each op’s expected dtype set.

shape: Arc<[u32]>

Logical shape in elements (not bytes). An empty vec is a scalar; a 2-element vec is a matrix; etc. Used for shape-mismatch detection at build-time.

Implementations

impl TensorRef

pub fn new(name: impl Into<Ident>, dtype: DataType, shape: Vec<u32>) -> Self

Construct an explicit TensorRef. Callers prefer the shape helpers below unless their shape is computed.

pub fn u32_1d(name: impl Into<Ident>, len: u32) -> Self

U32 1-D tensor convenience constructor.

pub fn f32_1d(name: impl Into<Ident>, len: u32) -> Self

F32 1-D tensor convenience constructor.

pub fn u32_2d(name: impl Into<Ident>, rows: u32, cols: u32) -> Self

U32 2-D tensor convenience constructor (rows × cols).

pub fn f16_1d(name: impl Into<Ident>, len: u32) -> Self

F16 1-D tensor convenience constructor.

pub fn f16_2d(name: impl Into<Ident>, rows: u32, cols: u32) -> Self

F16 2-D tensor convenience constructor (rows × cols).

pub fn f32_2d(name: impl Into<Ident>, rows: u32, cols: u32) -> Self

F32 2-D tensor convenience constructor (rows × cols).

pub fn element_count(&self) -> Option<u32>

Total element count. Returns None on overflow so builders can surface a structured error rather than silent wraparound.

pub fn name_str(&self) -> &str

Borrow the buffer name as &str - the form every IR builder still accepts. Lets Cat-A ops forward to underlying primitives while keeping the typed surface on the boundary.

Trait Implementations

impl Clone for TensorRef

fn clone(&self) -> TensorRef

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for TensorRef

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

Formats the value using the given formatter.

impl PartialEq for TensorRef

fn eq(&self, other: &TensorRef) -> 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 Eq for TensorRef

impl StructuralPartialEq for TensorRef