Struct CudaResidentStorage 

pub struct CudaResidentStorage { /* private fields */ }

Resident storage for a CUDA matmul result.

Implementations

impl CudaResidentStorage

pub fn new( site: Symbol, allocation: CudaAllocation, shape: Vec<usize>, dtype: Symbol, cells: Arc<[Value]>, ) -> Self

Builds resident CUDA storage around checked host-equivalent cells.

pub fn allocation(&self) -> &CudaAllocation

Returns the resident allocation evidence.

pub fn resident_tensor(&self) -> Option<Tensor>

Rebuilds a canonical tensor without counting a user readback.

Trait Implementations

impl TensorStorage for CudaResidentStorage

fn dtype(&self) -> &Symbol

The logical scalar domain of every cell.

fn len(&self) -> usize

The logical row-major cell count.

fn location(&self) -> TensorLocation

The current storage placement.

fn cell(&self, index: usize) -> Result<Value>

Observes one row-major cell.

fn materialize(&self) -> Result<Arc<dyn TensorStorage>>

Returns a host-observable form of this storage.

fn as_any(&self) -> &dyn Any

Exposes the concrete storage for safe downcasting by typed adapters.

fn is_empty(&self) -> bool

Whether the storage has no logical cells.