Convolution

The convolution primitive computes a forward, backward, or weight update for a batched convolution operation on 1D, 2D, or 3D spatial data with bias. More...

Functions
dnnl_status_t DNNL_API	dnnl_convolution_forward_desc_init (dnnl_convolution_desc_t *conv_desc, dnnl_prop_kind_t prop_kind, dnnl_alg_kind_t alg_kind, const dnnl_memory_desc_t *src_desc, const dnnl_memory_desc_t *weights_desc, const dnnl_memory_desc_t *bias_desc, const dnnl_memory_desc_t *dst_desc, const dnnl_dims_t strides, const dnnl_dims_t padding_l, const dnnl_dims_t padding_r)
	Initializes a convolution descriptor conv_desc for forward propagation using prop_kind (possible values are dnnl_forward_training and dnnl_forward_inference), alg_kind, memory descriptors, strides, padding_l, and padding_r. More...
dnnl_status_t DNNL_API	dnnl_dilated_convolution_forward_desc_init (dnnl_convolution_desc_t *conv_desc, dnnl_prop_kind_t prop_kind, dnnl_alg_kind_t alg_kind, const dnnl_memory_desc_t *src_desc, const dnnl_memory_desc_t *weights_desc, const dnnl_memory_desc_t *bias_desc, const dnnl_memory_desc_t *dst_desc, const dnnl_dims_t strides, const dnnl_dims_t dilates, const dnnl_dims_t padding_l, const dnnl_dims_t padding_r)
	Initializes a dilated convolution descriptor conv_desc for forward propagation using prop_kind (possible values are dnnl_forward_training and dnnl_forward_inference), alg_kind, memory descriptors, strides, dilates, padding_l, and padding_r. More...
dnnl_status_t DNNL_API	dnnl_convolution_backward_data_desc_init (dnnl_convolution_desc_t *conv_desc, dnnl_alg_kind_t alg_kind, const dnnl_memory_desc_t *diff_src_desc, const dnnl_memory_desc_t *weights_desc, const dnnl_memory_desc_t *diff_dst_desc, const dnnl_dims_t strides, const dnnl_dims_t padding_l, const dnnl_dims_t padding_r)
	Initializes a convolution descriptor conv_desc for backward propagation with respect to data using alg_kind, memory descriptors, strides, padding_l, and padding_r. More...
dnnl_status_t DNNL_API	dnnl_dilated_convolution_backward_data_desc_init (dnnl_convolution_desc_t *conv_desc, dnnl_alg_kind_t alg_kind, const dnnl_memory_desc_t *diff_src_desc, const dnnl_memory_desc_t *weights_desc, const dnnl_memory_desc_t *diff_dst_desc, const dnnl_dims_t strides, const dnnl_dims_t dilates, const dnnl_dims_t padding_l, const dnnl_dims_t padding_r)
	Initializes a dilated convolution descriptor conv_desc for backward propagation with respect to data using alg_kind, memory descriptors, strides, dilates padding_l, and padding_r. More...
dnnl_status_t DNNL_API	dnnl_convolution_backward_weights_desc_init (dnnl_convolution_desc_t *conv_desc, dnnl_alg_kind_t alg_kind, const dnnl_memory_desc_t *src_desc, const dnnl_memory_desc_t *diff_weights_desc, const dnnl_memory_desc_t *diff_bias_desc, const dnnl_memory_desc_t *diff_dst_desc, const dnnl_dims_t strides, const dnnl_dims_t padding_l, const dnnl_dims_t padding_r)
	Initializes a convolution descriptor conv_desc for backward propagation with respect to weights using alg_kind, memory descriptors, strides, padding_l, and padding_r. More...
dnnl_status_t DNNL_API	dnnl_dilated_convolution_backward_weights_desc_init (dnnl_convolution_desc_t *conv_desc, dnnl_alg_kind_t alg_kind, const dnnl_memory_desc_t *src_desc, const dnnl_memory_desc_t *diff_weights_desc, const dnnl_memory_desc_t *diff_bias_desc, const dnnl_memory_desc_t *diff_dst_desc, const dnnl_dims_t strides, const dnnl_dims_t dilates, const dnnl_dims_t padding_l, const dnnl_dims_t padding_r)
	Initializes a convolution descriptor conv_desc for backward propagation with respect to weights using alg_kind, memory descriptors, strides, dilates padding_l, and padding_r. More...

Detailed Description

The convolution primitive computes a forward, backward, or weight update for a batched convolution operation on 1D, 2D, or 3D spatial data with bias.

See also

Convolution in developer guide

Convolution in C++ API

Function Documentation

dnnl_convolution_forward_desc_init()

dnnl_status_t DNNL_API dnnl_convolution_forward_desc_init	(	dnnl_convolution_desc_t *	conv_desc,
		dnnl_prop_kind_t	prop_kind,
		dnnl_alg_kind_t	alg_kind,
		const dnnl_memory_desc_t *	src_desc,
		const dnnl_memory_desc_t *	weights_desc,
		const dnnl_memory_desc_t *	bias_desc,
		const dnnl_memory_desc_t *	dst_desc,
		const dnnl_dims_t	strides,
		const dnnl_dims_t	padding_l,
		const dnnl_dims_t	padding_r
	)

Initializes a convolution descriptor conv_desc for forward propagation using prop_kind (possible values are dnnl_forward_training and dnnl_forward_inference), alg_kind, memory descriptors, strides, padding_l, and padding_r.

In order to create a convolution without bias, bias_desc should either be NULL or point to a descriptor with memory format kind equal to dnnl_format_kind_undef.

Note

If padding_r is NULL, the padding is supposed to be symmetric.

Memory descriptors are allowed to be initialized with dnnl_format_kind_any value of format_kind.

Inputs:

• src (dnnl_query_src_md, 0)
• weights (dnnl_query_weights_md, 0)
• bias (dnnl_query_weights_md, 1), if created with bias

Outputs:

• dst (dnnl_query_dst_md, 0)

Examples:

cnn_inference_f32.c, and cpu_cnn_training_f32.c.

dnnl_dilated_convolution_forward_desc_init()

dnnl_status_t DNNL_API dnnl_dilated_convolution_forward_desc_init	(	dnnl_convolution_desc_t *	conv_desc,
		dnnl_prop_kind_t	prop_kind,
		dnnl_alg_kind_t	alg_kind,
		const dnnl_memory_desc_t *	src_desc,
		const dnnl_memory_desc_t *	weights_desc,
		const dnnl_memory_desc_t *	bias_desc,
		const dnnl_memory_desc_t *	dst_desc,
		const dnnl_dims_t	strides,
		const dnnl_dims_t	dilates,
		const dnnl_dims_t	padding_l,
		const dnnl_dims_t	padding_r
	)

Initializes a dilated convolution descriptor conv_desc for forward propagation using prop_kind (possible values are dnnl_forward_training and dnnl_forward_inference), alg_kind, memory descriptors, strides, dilates, padding_l, and padding_r.

In order to create a dilated convolution without bias, bias_desc should either be NULL or point to a descriptor with memory format kind equals dnnl_format_kind_undef.

Note

If padding_r is NULL, the padding is supposed to be symmetric.

Memory descriptors are allowed to be initialized with dnnl_format_kind_any value of format_kind.

Inputs:

• src (dnnl_query_src_md, 0)
• weights (dnnl_query_weights_md, 0)
• bias (dnnl_query_weights_md, 1), if created with bias

Outputs:

• dst (dnnl_query_dst_md, 0)

dnnl_convolution_backward_data_desc_init()

dnnl_status_t DNNL_API dnnl_convolution_backward_data_desc_init	(	dnnl_convolution_desc_t *	conv_desc,
		dnnl_alg_kind_t	alg_kind,
		const dnnl_memory_desc_t *	diff_src_desc,
		const dnnl_memory_desc_t *	weights_desc,
		const dnnl_memory_desc_t *	diff_dst_desc,
		const dnnl_dims_t	strides,
		const dnnl_dims_t	padding_l,
		const dnnl_dims_t	padding_r
	)

Initializes a convolution descriptor conv_desc for backward propagation with respect to data using alg_kind, memory descriptors, strides, padding_l, and padding_r.

Note

Memory descriptors are allowed to be initialized with dnnl_format_kind_any value of format_kind.

Inputs:

• diff_dst (dnnl_query_diff_dst_md, 0)
• weights (dnnl_query_weights_md, 0)

Outputs:

• diff_src (dnnl_query_diff_src_md, 0)

dnnl_dilated_convolution_backward_data_desc_init()

dnnl_status_t DNNL_API dnnl_dilated_convolution_backward_data_desc_init	(	dnnl_convolution_desc_t *	conv_desc,
		dnnl_alg_kind_t	alg_kind,
		const dnnl_memory_desc_t *	diff_src_desc,
		const dnnl_memory_desc_t *	weights_desc,
		const dnnl_memory_desc_t *	diff_dst_desc,
		const dnnl_dims_t	strides,
		const dnnl_dims_t	dilates,
		const dnnl_dims_t	padding_l,
		const dnnl_dims_t	padding_r
	)

Initializes a dilated convolution descriptor conv_desc for backward propagation with respect to data using alg_kind, memory descriptors, strides, dilates padding_l, and padding_r.

Note

Memory descriptors are allowed to be initialized with dnnl_format_kind_any value of format_kind.

Inputs:

• diff_dst (dnnl_query_diff_dst_md, 0)
• weights (dnnl_query_weights_md, 0)

Outputs:

• diff_src (dnnl_query_diff_src_md, 0)

dnnl_convolution_backward_weights_desc_init()

dnnl_status_t DNNL_API dnnl_convolution_backward_weights_desc_init	(	dnnl_convolution_desc_t *	conv_desc,
		dnnl_alg_kind_t	alg_kind,
		const dnnl_memory_desc_t *	src_desc,
		const dnnl_memory_desc_t *	diff_weights_desc,
		const dnnl_memory_desc_t *	diff_bias_desc,
		const dnnl_memory_desc_t *	diff_dst_desc,
		const dnnl_dims_t	strides,
		const dnnl_dims_t	padding_l,
		const dnnl_dims_t	padding_r
	)

Initializes a convolution descriptor conv_desc for backward propagation with respect to weights using alg_kind, memory descriptors, strides, padding_l, and padding_r.

Note

Memory descriptors are allowed to be initialized with dnnl_format_kind_any value of format_kind.

Inputs:

• src (dnnl_query_src_md, 0)
• diff_dst (dnnl_query_diff_dst_md, 0)

Outputs:

• diff_weights (dnnl_query_diff_weights_md, 0)
• diff_bias (dnnl_query_diff_weights_md, 1), if created with bias

Examples:

cpu_cnn_training_f32.c.

dnnl_dilated_convolution_backward_weights_desc_init()

dnnl_status_t DNNL_API dnnl_dilated_convolution_backward_weights_desc_init	(	dnnl_convolution_desc_t *	conv_desc,
		dnnl_alg_kind_t	alg_kind,
		const dnnl_memory_desc_t *	src_desc,
		const dnnl_memory_desc_t *	diff_weights_desc,
		const dnnl_memory_desc_t *	diff_bias_desc,
		const dnnl_memory_desc_t *	diff_dst_desc,
		const dnnl_dims_t	strides,
		const dnnl_dims_t	dilates,
		const dnnl_dims_t	padding_l,
		const dnnl_dims_t	padding_r
	)

Initializes a convolution descriptor conv_desc for backward propagation with respect to weights using alg_kind, memory descriptors, strides, dilates padding_l, and padding_r.

Note

Memory descriptors are allowed to be initialized with dnnl_format_kind_any value of format_kind.

Inputs:

• src (dnnl_query_src_md, 0)
• diff_dst (dnnl_query_diff_dst_md, 0)

Outputs:

• diff_weights (dnnl_query_diff_weights_md, 0)
• diff_bias (dnnl_query_diff_weights_md, 1), if created with bias