Intel(R) Math Kernel Library for Deep Neural Networks (Intel(R) MKL-DNN)  1.0.4
Performance library for Deep Learning
Classes | Typedefs | Enumerations
Operation descriptors

Classes

struct  mkldnn_convolution_desc_t
 A descriptor of a convolution operation. More...
 
struct  mkldnn_shuffle_desc_t
 A descriptor of a shuffle operation. More...
 
struct  mkldnn_eltwise_desc_t
 A descriptor of a element-wise operation. More...
 
struct  mkldnn_softmax_desc_t
 A descriptor of a Softmax operation. More...
 
struct  mkldnn_pooling_desc_t
 A descriptor of a pooling operation. More...
 
struct  mkldnn_lrn_desc_t
 A descriptor of a Local Response Normalization (LRN) operation. More...
 
struct  mkldnn_batch_normalization_desc_t
 A descriptor of a Batch Normalization operation. More...
 
struct  mkldnn_inner_product_desc_t
 A descriptor of an inner product operation. More...
 
struct  mkldnn_rnn_desc_t
 A descriptor for an RNN operation. More...
 

Typedefs

typedef void * mkldnn_op_desc_t
 A pointer to any of the operation descriptors.
 
typedef const void * const_mkldnn_op_desc_t
 A pointer to any of the operation descriptors (constant variant).
 
typedef mkldnn_convolution_desc_t mkldnn_deconvolution_desc_t
 A descriptor of a deconvolution operation.
 

Enumerations

enum  mkldnn_rnn_flags_t
 Flags for RNN cell.
 
enum  mkldnn_rnn_direction_t { mkldnn_unidirectional_left2right, mkldnn_unidirectional_right2left, mkldnn_bidirectional_concat, mkldnn_bidirectional_sum }
 A direction of RNN primitive execution. More...
 

Detailed Description

Enumeration Type Documentation

◆ mkldnn_rnn_direction_t

A direction of RNN primitive execution.

Enumerator
mkldnn_unidirectional_left2right 

Unidirectional execution of RNN primitive from left to right.

mkldnn_unidirectional_right2left 

Unidirectional execution of RNN primitive from right to left.

mkldnn_bidirectional_concat 

Bidirectional execution of RNN primitive with concatenation of the results.

mkldnn_bidirectional_sum 

Bidirectional execution of RNN primitive with summation of the results.