Intel® oneAPI Deep Neural Network Developer Guide and Reference
A newer version of this document is available. Customers should click here to go to the newest version.
Tensor
Overview
Tensor is an abstraction for multi-dimensional input and output data needed in the execution of a compiled partition. More…
// typedefs
typedef struct dnnl_graph_tensor* dnnl_graph_tensor_t;
typedef const struct dnnl_graph_tensor* const_dnnl_graph_tensor_t;
// classes
class dnnl::graph::tensor;
// global functions
dnnl_status_t DNNL_API dnnl_graph_tensor_create(
dnnl_graph_tensor_t* tensor,
const dnnl_graph_logical_tensor_t* logical_tensor,
dnnl_engine_t engine,
void* handle
);
dnnl_status_t DNNL_API dnnl_graph_tensor_destroy(dnnl_graph_tensor_t tensor);
dnnl_status_t DNNL_API dnnl_graph_tensor_get_data_handle(
const_dnnl_graph_tensor_t tensor,
void** handle
);
dnnl_status_t DNNL_API dnnl_graph_tensor_set_data_handle(
dnnl_graph_tensor_t tensor,
void* handle
);
dnnl_status_t DNNL_API dnnl_graph_tensor_get_engine(
const_dnnl_graph_tensor_t tensor,
dnnl_engine_t* engine
);
dnnl_status_t DNNL_API dnnl_graph_tensor_get_logical_tensor(
const_dnnl_graph_tensor_t tensor,
dnnl_graph_logical_tensor_t* logical_tensor
);Detailed Documentation
Tensor is an abstraction for multi-dimensional input and output data needed in the execution of a compiled partition.
A tensor object encapsulates a handle to a memory buffer allocated on a specific engine and a logical tensor which describes the dimensions, elements data type, and memory layout.
Typedefs
typedef struct dnnl_graph_tensor* dnnl_graph_tensor_tA tensor handle.
typedef const struct dnnl_graph_tensor* const_dnnl_graph_tensor_tA constant tensor handle.
Global Functions
dnnl_status_t DNNL_API dnnl_graph_tensor_create(
dnnl_graph_tensor_t* tensor,
const dnnl_graph_logical_tensor_t* logical_tensor,
dnnl_engine_t engine,
void* handle
)Creates a tensor with logical tensor, engine, and data handle.
Parameters:
tensor |
Output tensor. |
logical_tensor |
Description for this tensor. |
engine |
Engine to use. |
handle |
Handle of the memory buffer to use as an underlying storage.
|
Returns:
dnnl_success on success or a status describing the error otherwise.
dnnl_status_t DNNL_API dnnl_graph_tensor_destroy(dnnl_graph_tensor_t tensor)Destroys a tensor.
Parameters:
tensor |
The tensor to be destroyed. |
Returns:
dnnl_success on success or a status describing the error otherwise.
dnnl_status_t DNNL_API dnnl_graph_tensor_get_data_handle(
const_dnnl_graph_tensor_t tensor,
void** handle
)Gets the data handle of a tensor.
Parameters:
tensor |
The input tensor. |
handle |
Pointer to the data of input tensor. |
Returns:
dnnl_success on success or a status describing the error otherwise.
dnnl_status_t DNNL_API dnnl_graph_tensor_set_data_handle(
dnnl_graph_tensor_t tensor,
void* handle
)Set data handle for a tensor.
Parameters:
tensor |
The input tensor. |
handle |
New data handle for tensor. |
Returns:
dnnl_success on success or a status describing the error otherwise.
dnnl_status_t DNNL_API dnnl_graph_tensor_get_engine(
const_dnnl_graph_tensor_t tensor,
dnnl_engine_t* engine
)Returns the engine of a tensor object.
Parameters:
tensor |
The input tensor. |
engine |
Output engine on which the tensor is located. |
Returns:
dnnl_success on success or a status describing the error otherwise.
dnnl_status_t DNNL_API dnnl_graph_tensor_get_logical_tensor(
const_dnnl_graph_tensor_t tensor,
dnnl_graph_logical_tensor_t* logical_tensor
)Returns the logical tensor of a tensor object.
Parameters:
tensor |
The input tensor. |
logical_tensor |
Output logical tensor of the tensor object. |
Returns:
dnnl_success on success or a status describing the error otherwise.