Struct CudaTensorInfo 

pub struct CudaTensorInfo {
    pub device_id: i32,
    pub shape: Vec<usize>,
    pub dtype: DLDataType,
    pub byte_offset: u64,
    pub device_type_code: i32,
}

Metadata extracted from a CUDA-resident DLPack tensor.

Accessing this struct does not copy tensor data from GPU to CPU. It records only the shape, dtype, device index, and byte offset that are safe to read from the capsule header.

Fields

device_id: i32

Zero-based index of the CUDA device (e.g., 0 for the first GPU).

shape: Vec<usize>

Tensor dimensions in row-major (C) order.

dtype: DLDataType

DLPack element data-type descriptor.

byte_offset: u64

Byte offset from the data pointer to the first element.

device_type_code: i32

Raw device-type code (2 = CUDA, 10 = ROCm, etc.).

Implementations

impl CudaTensorInfo

pub fn numel(&self) -> usize

Return the total number of elements (product of shape dimensions).

pub fn dtype_bits(&self) -> u8

Return the element bit-width.

pub fn device_str(&self) -> String

Return a human-readable device string (e.g. "cuda:0").

Trait Implementations

impl Clone for CudaTensorInfo

fn clone(&self) -> CudaTensorInfo

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for CudaTensorInfo

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

Formats the value using the given formatter.