blob: ce0cf53fdff38c1713ffd88c1436bdb4e3bd8a73 [file] [log] [blame]
Anthony Barbier6ff3b192017-09-04 18:44:23 +01001/*
Diego Lopez Recas35ceeb22017-12-04 18:56:10 +00002 * Copyright (c) 2016-2018 ARM Limited.
Anthony Barbier6ff3b192017-09-04 18:44:23 +01003 *
4 * SPDX-License-Identifier: MIT
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to
8 * deal in the Software without restriction, including without limitation the
9 * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
10 * sell copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
12 *
13 * The above copyright notice and this permission notice shall be included in all
14 * copies or substantial portions of the Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
19 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
22 * SOFTWARE.
23 */
24#ifndef __ARM_COMPUTE_ITENSORINFO_H__
25#define __ARM_COMPUTE_ITENSORINFO_H__
26
27#include "arm_compute/core/Coordinates.h"
28#include "arm_compute/core/Strides.h"
29#include "arm_compute/core/TensorShape.h"
30#include "arm_compute/core/Types.h"
31#include "arm_compute/core/Utils.h"
Georgios Pinitas283c1792017-11-10 18:14:06 +000032#include "arm_compute/core/utils/misc/ICloneable.h"
Georgios Pinitasd8734b52017-12-22 15:27:52 +000033#include "arm_compute/core/utils/misc/Utility.h"
Anthony Barbier6ff3b192017-09-04 18:44:23 +010034
35#include <cstddef>
36
37namespace arm_compute
38{
39/** Store the tensor's metadata */
Georgios Pinitas283c1792017-11-10 18:14:06 +000040class ITensorInfo : public misc::ICloneable<ITensorInfo>
Anthony Barbier6ff3b192017-09-04 18:44:23 +010041{
42public:
43 /** Default virtual destructor */
44 virtual ~ITensorInfo() = default;
45 /** Set the data type to the specified value.
46 *
47 * @warning This resets the format to UNKNOWN.
48 *
49 * @param[in] data_type The new data type.
Georgios Pinitas283c1792017-11-10 18:14:06 +000050 *
51 * @return Reference to this ITensorInfo object
Anthony Barbier6ff3b192017-09-04 18:44:23 +010052 */
Georgios Pinitas283c1792017-11-10 18:14:06 +000053 virtual ITensorInfo &set_data_type(DataType data_type) = 0;
Anthony Barbier6ff3b192017-09-04 18:44:23 +010054 /** Set the number of channels to the specified value.
55 *
56 * @warning This resets the format to UNKNOWN.
57 *
58 * @param[in] num_channels New number of channels.
Georgios Pinitas283c1792017-11-10 18:14:06 +000059 *
60 * @return Reference to this ITensorInfo object
Anthony Barbier6ff3b192017-09-04 18:44:23 +010061 */
Georgios Pinitas283c1792017-11-10 18:14:06 +000062 virtual ITensorInfo &set_num_channels(int num_channels) = 0;
Anthony Barbier6ff3b192017-09-04 18:44:23 +010063 /** Set the format of an already initialized tensor.
64 *
65 * @note If the data type has already been configured (i.e. not UNKNOWN) it
66 * must match the new format. If data type hasn't been configured it will
67 * be based on the format.
68 *
69 * @param[in] format Single-plane format of the tensor.
Georgios Pinitas283c1792017-11-10 18:14:06 +000070 *
71 * @return Reference to this ITensorInfo object
Anthony Barbier6ff3b192017-09-04 18:44:23 +010072 */
Georgios Pinitas283c1792017-11-10 18:14:06 +000073 virtual ITensorInfo &set_format(Format format) = 0;
Anthony Barbier6ff3b192017-09-04 18:44:23 +010074 /** Set the shape of an already initialized tensor.
75 *
76 * @warning Changing the shape requires to recompute the strides and is
77 * therefore only possible if the tensor hasn't been allocated yet.
78 *
79 * @param[in] shape New tensor shape.
Georgios Pinitas283c1792017-11-10 18:14:06 +000080 *
81 * @return Reference to this ITensorInfo object
Anthony Barbier6ff3b192017-09-04 18:44:23 +010082 */
Diego Lopez Recas35ceeb22017-12-04 18:56:10 +000083 virtual ITensorInfo &set_tensor_shape(const TensorShape &shape) = 0;
Anthony Barbier6ff3b192017-09-04 18:44:23 +010084 /** Set the fixed point position to the specified value
85 *
86 * @warning The fixed point position must be set once the data type has been configured
87 *
88 * @param[in] fixed_point_position The new fixed point position
Georgios Pinitas283c1792017-11-10 18:14:06 +000089 *
90 * @return Reference to this ITensorInfo object
Anthony Barbier6ff3b192017-09-04 18:44:23 +010091 */
Georgios Pinitas283c1792017-11-10 18:14:06 +000092 virtual ITensorInfo &set_fixed_point_position(int fixed_point_position) = 0;
93 /** Set the quantization settings (scale and offset) of the tensor.
Anthony Barbierf202e502017-11-23 18:02:04 +000094 *
95 * @param[in] quantization_info QuantizationInfo containing the scale and offset
96 *
97 * @return Reference to this ITensorInfo object
98 */
Diego Lopez Recas35ceeb22017-12-04 18:56:10 +000099 virtual ITensorInfo &set_quantization_info(const QuantizationInfo &quantization_info) = 0;
Isabella Gottardid17a6772018-02-27 17:41:55 +0000100 /** Set the data layout of the tensor.
101 *
102 * @param[in] data_layout DataLayout containing the layout data information.
103 *
104 * @return Reference to this ITensorInfo object
105 */
106 virtual ITensorInfo &set_data_layout(const DataLayout &data_layout) = 0;
Georgios Pinitas30902ed2017-11-14 15:32:57 +0000107 /** Resets the padding settings of the tensor.
108 *
109 * @return Reference to this ITensorInfo object
110 */
111 virtual ITensorInfo &reset_padding() = 0;
Anthony Barbier6ff3b192017-09-04 18:44:23 +0100112 /** Update the offset to the first element and the strides to automatically computed values.
113 *
114 * @note The padding used by this method is really conservative so that the tensor can be used for most functions.
115 *
116 * @return True if the strides or the offset to the first element have changed.
117 */
118 virtual bool auto_padding() = 0;
119 /** Update the offset to the first element, the strides and the total size.
120 *
121 * @note This function can only increase the offset, strides and total size.
122 *
123 * @param[in] padding Padding around the XY plane in number of elements.
124 *
125 * @return True if the strides, offset and total size have changed.
126 */
127 virtual bool extend_padding(const PaddingSize &padding) = 0;
128 /** Return the size of the requested dimension
129 *
130 * @param[in] index Index of the dimension
131 *
132 * @return Dimension of the requested dimension
133 */
134 virtual size_t dimension(size_t index) const = 0;
Isabella Gottardid56e7702018-02-28 14:29:36 +0000135 /** Return the size of the requested data layout dimension
136 *
137 * @param[in] dimension DataLayoutDimension of the dimension
138 *
139 * @return Dimension of the requested dimension
140 */
141 virtual size_t dimension(DataLayoutDimension dimension) const = 0;
Anthony Barbier6ff3b192017-09-04 18:44:23 +0100142 /** The strides in bytes for accessing each dimension of the tensor
143 *
144 * @return Strides in bytes for each tensor dimension
145 */
146 virtual const Strides &strides_in_bytes() const = 0;
147 /** The offset from the beginning of the memory allocation to the first element of the tensor.
148 * This can be used to access efficiently elements in a 2D tensor
149 *
150 * @return The offset in bytes to access the first element of the tensor.
151 */
152 virtual size_t offset_first_element_in_bytes() const = 0;
153 /** The offset in bytes from the beginning of the memory allocation to access the element at position (x, y, z ...)
154 *
155 * @param[in] pos Vector with the coordinates of the element to access.
156 * The size of this vector must be equal to the number of dimensions of the tensor
157 *
158 * @return Offset in bytes from the beginning of the memory allocation to access the element (x, y, z, ...)
159 */
160 virtual size_t offset_element_in_bytes(const Coordinates &pos) const = 0;
161 /** Fixed point position used when the tensor data type is QS8 or QS16
162 *
163 * @return The fixed point position that expresses the number of bits for the fractional part of the number
164 */
165 virtual int fixed_point_position() const = 0;
166 /** Element size in bytes calculated as data_size() * num_channels()
167 *
168 * @return The size of one element in bytes
169 */
170 virtual size_t element_size() const = 0;
171 /** The number of dimensions of the tensor (rank)
172 *
173 * @return The number of dimensions of the tensor (rank)
174 */
175 virtual size_t num_dimensions() const = 0;
176 /** The number of channels for each tensor element
177 *
178 * @return The number of channels for each tensor element
179 */
180 virtual size_t num_channels() const = 0;
181 /** Size for each dimension of the tensor
182 *
183 * @return A vector with the size for each dimension of the tensor
184 */
185 virtual const TensorShape &tensor_shape() const = 0;
186 /** Data type used for each element of the tensor
187 *
188 * @return Tensor data type
189 */
190 virtual DataType data_type() const = 0;
191 /** Colour format of the image
192 *
193 * @return Colour format of the image
194 */
195 virtual Format format() const = 0;
196 /** Returns the total size of the tensor in bytes.
197 *
198 * @return Total size of the tensor in bytes.
199 */
200 virtual size_t total_size() const = 0;
201 /** Padding of tensor.
202 *
203 * @return Padding.
204 */
205 virtual PaddingSize padding() const = 0;
206 /** Checks if the tensor has been allocated with padding or not.
207 *
208 * @return True if padding is allocated in the tensor, otherwise false.
209 */
210 virtual bool has_padding() const = 0;
211 /** Flag indicating whether the size of the tensor can be changed.
212 *
213 * @return True if the tensor size can be changed.
214 */
215 virtual bool is_resizable() const = 0;
216 /** Set the flag whether the tensor size can be changed.
217 *
218 * @param[in] is_resizable Flag that marks the tensor if it can be changed or not.
Georgios Pinitasa3b1b462017-11-16 19:24:39 +0000219 *
220 * @return Reference to this ITensorInfo object
Anthony Barbier6ff3b192017-09-04 18:44:23 +0100221 */
Georgios Pinitasa3b1b462017-11-16 19:24:39 +0000222 virtual ITensorInfo &set_is_resizable(bool is_resizable) = 0;
Anthony Barbier6ff3b192017-09-04 18:44:23 +0100223 /** Valid region of the tensor. All elements in the valid region have defined values, i.e. are not undefined.
224 *
225 * @return The valid region.
226 */
227 virtual ValidRegion valid_region() const = 0;
228 /** Set the valid region of the tensor.
229 *
230 * @param[in] valid_region Valid region to set.
231 */
Diego Lopez Recas35ceeb22017-12-04 18:56:10 +0000232 virtual void set_valid_region(const ValidRegion &valid_region) = 0;
Michel Iwaniec00633802017-10-12 14:14:15 +0100233
234 /** Get the quantization settings (scale and offset) of the tensor.
235 *
236 * @return A QuantizationInfo containing the scale and offset.
237 */
238 virtual QuantizationInfo quantization_info() const = 0;
Isabella Gottardid17a6772018-02-27 17:41:55 +0000239 /** Get the data layout of the tensor.
240 *
241 * @return A DataLayout containing the layout data information.
242 */
243 virtual DataLayout data_layout() const = 0;
Diego Lopez Recas0021d752017-12-18 14:42:56 +0000244
245 /** If infos are broadcast compatible tensor info's, return the broadcasted shape and the intersection of
246 * the broadcasted valid regions of the tensors.
247 *
248 * Two tensor info's are broadcast compatible if their shapes are broadcast compatible.
249 *
250 * Two tensor shapes are broadcast compatible if for each dimension, they're equal or one of them is 1.
251 *
252 * If two shapes are compatible, each dimension in the broadcasted shape is the max of the original dimensions.
253 *
254 * @param[in] infos Tensor info's.
255 *
256 * @return The broadcasted shape and valid region, or an empty shape and valid region if the info's are
257 * not broadcast compatible.
258 */
259 template <typename... Infos>
260 static std::pair<TensorShape, ValidRegion> broadcast_shape_and_valid_region(const Infos &... infos)
261 {
262 TensorShape bc_shape = TensorShape::broadcast_shape(infos.tensor_shape()...);
263 ValidRegion bc_valid_region{ Coordinates(), bc_shape };
264
265 auto broadcast_valid_region = [&bc_valid_region](const ITensorInfo & info)
266 {
267 if(info.num_dimensions() != 0)
268 {
269 for(size_t d = 0; d < bc_valid_region.shape.num_dimensions(); ++d)
270 {
271 const bool is_broadcast = (info.tensor_shape()[d] == 1);
272
273 const int anchor_max = std::max(bc_valid_region.anchor[d], info.valid_region().anchor[d]);
274 const size_t valid_min = std::min(bc_valid_region.shape[d], info.valid_region().shape[d]);
275
276 if(!is_broadcast || (valid_min == 0))
277 {
278 bc_valid_region.anchor.set(d, anchor_max);
279 bc_valid_region.shape.set(d, valid_min);
280 }
281 }
282 }
283 };
284
285 utility::for_each(broadcast_valid_region, infos...);
286
287 return std::pair<TensorShape, ValidRegion>(bc_shape, bc_valid_region);
288 }
Anthony Barbier6ff3b192017-09-04 18:44:23 +0100289};
290}
291#endif /*__ARM_COMPUTE_TENSORINFO_H__ */