Deep Neural Network Library (DNNL)  1.3.0
Performance library for Deep Learning
Classes | Functions

A primitive to shuffle tensor data along an axis. More...

Classes

struct  dnnl::shuffle_forward
 Shuffle forward propagation primitive. More...
 
struct  dnnl::shuffle_backward
 Shuffle backward propagation primitive. More...
 
struct  dnnl_shuffle_desc_t
 A descriptor of a shuffle operation. More...
 

Functions

dnnl_status_t DNNL_API dnnl_shuffle_forward_desc_init (dnnl_shuffle_desc_t *shuffle_desc, dnnl_prop_kind_t prop_kind, const dnnl_memory_desc_t *data_desc, int axis, dnnl_dim_t group_size)
 Initializes a descriptor for shuffle forward propagation primitive. More...
 
dnnl_status_t DNNL_API dnnl_shuffle_backward_desc_init (dnnl_shuffle_desc_t *shuffle_desc, const dnnl_memory_desc_t *diff_data_desc, int axis, dnnl_dim_t group_size)
 Initializes a descriptor for shuffle backward propagation primitive. More...
 

Detailed Description

A primitive to shuffle tensor data along an axis.

See also
Shuffle in developer guide

Function Documentation

◆ dnnl_shuffle_forward_desc_init()

dnnl_status_t DNNL_API dnnl_shuffle_forward_desc_init ( dnnl_shuffle_desc_t shuffle_desc,
dnnl_prop_kind_t  prop_kind,
const dnnl_memory_desc_t data_desc,
int  axis,
dnnl_dim_t  group_size 
)

Initializes a descriptor for shuffle forward propagation primitive.

Inputs:

Outputs:

Parameters
shuffle_descOutput descriptor for a shuffle primitive.
prop_kindPropagation kind. Possible values are dnnl_forward_training and dnnl_forward_inference.
data_descSource and destination memory descriptor.
axisThe axis along which the data is shuffled.
group_sizeShuffle group size.
Returns
dnnl_success on success and a status describing the error otherwise.

◆ dnnl_shuffle_backward_desc_init()

dnnl_status_t DNNL_API dnnl_shuffle_backward_desc_init ( dnnl_shuffle_desc_t shuffle_desc,
const dnnl_memory_desc_t diff_data_desc,
int  axis,
dnnl_dim_t  group_size 
)

Initializes a descriptor for shuffle backward propagation primitive.

Inputs:

Outputs:

Parameters
shuffle_descOutput descriptor for a shuffle primitive.
diff_data_descDiff source and diff destination memory descriptor.
axisThe axis along which the data is shuffled.
group_sizeShuffle group size.
Returns
dnnl_success on success and a status describing the error otherwise.