vt::CuTensor

template<typename T, size_t N>
class CuTensor

CuTensor class. It is used to adapt a Tensor to be used in CUDA kernel.

Template Parameters:

• T – Data type of the tensor.

• N – Number of dimensions of the tensor.

Public Functions

inline __host__ CuTensor(Tensor<T, N> tensor)

Construct a new CuTensor object from a Tensor.

Parameters:

tensor – The tensor to be adapted.

inline __host__ __device__ CuTensor (T *data, const size_t *shape, const size_t *strides, const size_t start, const size_t size, const Order order, const bool contiguous)

Construct a new CuTensor object from a raw pointer, shape, strides, start index, and contiguous flag.

Parameters:

• data – The raw pointer to the data.

• shape – The shape of the tensor.

• strides – The strides of the tensor.

• start – The start index of the tensor.

• size – The size of the tensor.

• order – The order of the tensor.

• contiguous – The contiguous flag of the tensor.

template<typename... Args> inline __host__ __device__ size_t get_tensor_index (Args... args) const

Helper function to get the tensor index. (e.x. tensor(0, 1, 2))

Parameters:

args – The indices of the tensor.

Template Parameters:

Args – Variadic template for mulitple arguments.

Returns:

size_t: The index of the tensor.

template<typename... Args> inline __host__ __device__ T & operator() (Args... args)

Return the data given the indices.

Parameters:

args – The indices of the tensor.

Template Parameters:

Args – Variadic template for mulitple arguments.

Returns:

T: The data of the tensor.

template<typename... Args> inline __host__ __device__ const T & operator() (Args... args) const

Return the data given the indices.

Parameters:

args – The indices of the tensor.

Template Parameters:

Args – Variadic template for mulitple arguments.

Returns:

T: The data of the tensor.

inline __host__ __device__ T & operator[] (size_t n)

Return the data given the iterative index.

Parameters:

n – The 1D index of the tensor.

Returns:

T: The data of the tensor.

inline __host__ __device__ const T & operator[] (size_t n) const

Return the data given the iterative index.

Parameters:

n – The 1D index of the tensor.

Returns:

T: The data of the tensor.