#include <dnn_backend_native.h>

Data Fields
int32_t	dims [4]
	there are two memory layouts, NHWC or NCHW, so we use dims, dims[0] is Number. More...

DNNOperandType	type
	input/output/intermediate operand of the network More...

DNNDataType	data_type
	support different kinds of data type such as float, half float, int8 etc, first support float now. More...

int8_t	isNHWC
	NHWC if 1, otherwise NCHW. More...

char	name [128]
	to avoid possible memory leak, do not use char *name More...

void *	data
	data pointer with data length in bytes. More...

int32_t	length

int32_t	usedNumbersLeft

Detailed Description

Definition at line 68 of file dnn_backend_native.h.

Field Documentation

◆ dims

int32_t DnnOperand::dims[4]

there are two memory layouts, NHWC or NCHW, so we use dims, dims[0] is Number.

Definition at line 73 of file dnn_backend_native.h.

Referenced by convert_from_tensorflow.Operand::__str__(), dnn_execute_layer_conv2d_thread(), execute_model_native(), ff_calculate_operand_data_length(), ff_calculate_operand_dims_count(), ff_dnn_execute_layer_avg_pool(), ff_dnn_execute_layer_conv2d(), ff_dnn_execute_layer_dense(), ff_dnn_execute_layer_depth2space(), ff_dnn_execute_layer_pad(), ff_dnn_load_model_native(), get_input_native(), test(), test_broadcast_input0(), test_broadcast_input1(), test_no_broadcast(), test_with_mode_constant(), test_with_mode_reflect(), test_with_mode_symmetric(), test_with_same(), test_with_same_dilate(), and test_with_valid().

◆ type

DNNOperandType DnnOperand::type

input/output/intermediate operand of the network

Definition at line 78 of file dnn_backend_native.h.

Referenced by execute_model_native(), ff_dnn_load_model_native(), and get_input_native().

◆ data_type

DNNDataType DnnOperand::data_type

support different kinds of data type such as float, half float, int8 etc, first support float now.

Definition at line 84 of file dnn_backend_native.h.

Referenced by execute_model_native(), ff_dnn_execute_layer_avg_pool(), ff_dnn_execute_layer_conv2d(), ff_dnn_execute_layer_dense(), ff_dnn_execute_layer_depth2space(), ff_dnn_execute_layer_pad(), ff_dnn_load_model_native(), and get_input_native().

◆ isNHWC

int8_t DnnOperand::isNHWC

NHWC if 1, otherwise NCHW.

let's first support NHWC only, this flag is for extensive usage.

Definition at line 90 of file dnn_backend_native.h.

Referenced by ff_dnn_load_model_native().

◆ name

char DnnOperand::name[128]

to avoid possible memory leak, do not use char *name

Definition at line 95 of file dnn_backend_native.h.

Referenced by convert_from_tensorflow.Operand::__str__(), execute_model_native(), ff_dnn_load_model_native(), and get_input_native().

◆ data

void* DnnOperand::data

data pointer with data length in bytes.

usedNumbersLeft is only valid for intermediate operand, it means how many layers still depend on this operand, todo: the memory can be reused when usedNumbersLeft is zero.

Definition at line 103 of file dnn_backend_native.h.

Referenced by dnn_execute_layer_conv2d_thread(), execute_model_native(), ff_dnn_execute_layer_avg_pool(), ff_dnn_execute_layer_conv2d(), ff_dnn_execute_layer_dense(), ff_dnn_execute_layer_depth2space(), ff_dnn_execute_layer_pad(), ff_dnn_free_model_native(), math_binary_commutative(), math_binary_not_commutative(), test(), test_broadcast_input0(), test_broadcast_input1(), test_no_broadcast(), test_with_mode_constant(), test_with_mode_reflect(), test_with_mode_symmetric(), test_with_same(), test_with_same_dilate(), and test_with_valid().