xref: /aosp_15_r20/external/ComputeLibrary/src/core/NEON/kernels/NECropKernel.cpp (revision c217d954acce2dbc11938adb493fc0abd69584f3)
1 /*
2  * Copyright (c) 2019-2022 Arm Limited.
3  *
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 #include "src/core/NEON/kernels/NECropKernel.h"
25 
26 #include "arm_compute/core/ITensor.h"
27 #include "arm_compute/core/TensorInfo.h"
28 #include "arm_compute/core/Types.h"
29 #include "arm_compute/core/Window.h"
30 #include "arm_compute/core/utils/helpers/tensor_transform.h"
31 #include "arm_compute/core/utils/misc/ShapeCalculator.h"
32 #include "src/core/CPP/Validate.h"
33 #include "src/core/NEON/wrapper/wrapper.h"
34 #include "src/core/common/Registrars.h"
35 #include "src/core/helpers/AutoConfiguration.h"
36 #include "src/core/helpers/WindowHelpers.h"
37 #include "src/core/utils/helpers/bit_ops.h"
38 #include "src/cpu/kernels/crop/list.h"
39 
40 namespace arm_compute
41 {
42 namespace
43 {
44 struct CropSelectorData
45 {
46     DataType dt;
47 };
48 
49 using CropSelectorPtr = std::add_pointer<bool(const CropSelectorData &data)>::type;
50 using CropUKernelPtr  = std::add_pointer<void(const ITensor *, const ITensor *, float *, Coordinates, int32_t, int32_t, int32_t, bool, bool)>::type;
51 
52 struct CropUKernel
53 {
54     const char           *name;
55     const CropSelectorPtr is_selected;
56     CropUKernelPtr        ukernel;
57 };
58 
59 static const CropUKernel available_kernels[] =
60 {
61     {
62         "fp16_neon_crop",
__anon06bebd720202() 63         [](const CropSelectorData & data) { return data.dt == DataType::F16; },
64         REGISTER_FP16_NEON(arm_compute::cpu::fp16_in_bounds_crop_window)
65     },
66     {
67         "f32_neon_crop",
__anon06bebd720302() 68         [](const CropSelectorData & data) { return data.dt == DataType::F32; },
69         REGISTER_FP32_NEON(arm_compute::cpu::fp32_in_bounds_crop_window)
70     },
71     {
72         "u8_neon_crop",
__anon06bebd720402() 73         [](const CropSelectorData & data) { return data.dt == DataType::U8; },
74         REGISTER_INTEGER_NEON(arm_compute::cpu::u8_in_bounds_crop_window)
75     },
76     {
77         "u16_neon_crop",
__anon06bebd720502() 78         [](const CropSelectorData & data) { return data.dt == DataType::U16; },
79         REGISTER_INTEGER_NEON(arm_compute::cpu::u16_in_bounds_crop_window)
80     },
81     {
82         "u32_neon_crop",
__anon06bebd720602() 83         [](const CropSelectorData & data) { return data.dt == DataType::U32; },
84         REGISTER_INTEGER_NEON(arm_compute::cpu::u32_in_bounds_crop_window)
85     },
86     {
87         "s8_neon_crop",
__anon06bebd720702() 88         [](const CropSelectorData & data) { return data.dt == DataType::S8; },
89         REGISTER_INTEGER_NEON(arm_compute::cpu::s8_in_bounds_crop_window)
90     },
91     {
92         "s16_neon_crop",
__anon06bebd720802() 93         [](const CropSelectorData & data) { return data.dt == DataType::S16; },
94         REGISTER_INTEGER_NEON(arm_compute::cpu::s16_in_bounds_crop_window)
95     },
96     {
97         "s32_neon_crop",
__anon06bebd720902() 98         [](const CropSelectorData & data) { return data.dt == DataType::S32; },
99         REGISTER_INTEGER_NEON(arm_compute::cpu::s32_in_bounds_crop_window)
100     },
101 };
102 
103 /** Micro-kernel selector
104  *
105  * @param[in] data Selection data passed to help pick the appropriate micro-kernel
106  *
107  * @return A matching micro-kernel else nullptr
108  */
get_implementation(const CropSelectorData & data)109 const CropUKernel *get_implementation(const CropSelectorData &data)
110 {
111     for(const auto &uk : available_kernels)
112     {
113         if(uk.is_selected(data))
114         {
115             return &uk;
116         }
117     }
118 
119     return nullptr;
120 }
121 
out_of_bounds_crop_window(const ITensor * output,float * output_ptr,float extrapolation_value,int32_t window_step_x,int32_t output_width_start,int32_t output_width_limit)122 inline void out_of_bounds_crop_window(const ITensor *output, float *output_ptr, float extrapolation_value,
123                                       int32_t window_step_x, int32_t output_width_start, int32_t output_width_limit)
124 {
125     auto    in               = wrapper::vdup_n(extrapolation_value, wrapper::traits::vector_128_tag());
126     int32_t x                = 0;
127     int32_t limit            = (output_width_limit - output_width_start) * static_cast<int32_t>(output->info()->dimension(0));
128     float *output_start_ptr = output_ptr + output_width_start * output->info()->dimension(0);
129     for(; x <= limit - window_step_x; x += window_step_x)
130     {
131         wrapper::vstore(output_start_ptr + x, in);
132     }
133     for(; x < limit; ++x)
134     {
135         *(output_start_ptr + x) = extrapolation_value;
136     }
137 }
138 
execute_window(const ITensor * input,const ITensor * output,Coordinates input_offset,float extrapolation_value,const std::array<uint32_t,2> & rows_out_of_bounds,const std::array<uint32_t,2> & cols_out_of_bounds,NECropKernel::InBoundsCropFunction * in_bounds_crop_function,bool is_height_flipped,bool has_cols_in_bounds,bool has_cols_out_of_bounds_before,bool has_cols_out_of_bounds_after,bool input_has_single_channel,bool is_width_flipped)139 inline void execute_window(const ITensor *input, const ITensor *output, Coordinates input_offset, float extrapolation_value,
140                            const std::array<uint32_t, 2> &rows_out_of_bounds, const std::array<uint32_t, 2> &cols_out_of_bounds, NECropKernel::InBoundsCropFunction *in_bounds_crop_function,
141                            bool is_height_flipped, bool has_cols_in_bounds, bool has_cols_out_of_bounds_before, bool has_cols_out_of_bounds_after, bool input_has_single_channel, bool is_width_flipped)
142 {
143     // Output is always float.
144     const int window_step_x = 16 / sizeof(float);
145     auto     *output_ptr    = reinterpret_cast<float *>(output->buffer());
146     //  Output window:
147     //  --------------------------------
148     //  |          Out of bounds       |
149     //  |          rows before         |
150     //  |------------------------------|
151     //  | Out of | In         | Out of |
152     //  | bounds | bounds     | bounds |
153     //  | cols   | elements   | cols   |
154     //  | before | copied     | after  |
155     //  |        | from input |        |
156     //  --------------------------------
157     //  |        Out of bounds         |
158     //  |        rows after            |
159     //  |------------------------------|
160     // Fill all output rows that have no elements that are within the input bounds with the extrapolation value.
161     // First for the rows before the in bounds rows.
162     out_of_bounds_crop_window(output, output_ptr, extrapolation_value, window_step_x, 0, rows_out_of_bounds[0] * output->info()->dimension(1));
163     output_ptr += rows_out_of_bounds[0] * output->info()->dimension(1) * output->info()->dimension(0);
164     // Iterate through each row that has any elements within the input bounds.
165     for(uint32_t row = rows_out_of_bounds[0]; static_cast<int32_t>(row) < static_cast<int32_t>(output->info()->dimension(2) - rows_out_of_bounds[1]);
166         ++row, is_height_flipped ? --input_offset[2] : ++input_offset[2])
167     {
168         // Fill all elements in the row that are out of bounds with the extrapolation value.
169         // First for the elements before the in bounds elements.
170         if(has_cols_out_of_bounds_before)
171         {
172             out_of_bounds_crop_window(output, output_ptr, extrapolation_value, window_step_x, 0, cols_out_of_bounds[0]);
173         }
174         // Copy all elements within the input bounds from the input tensor.
175         if(has_cols_in_bounds)
176         {
177             (*in_bounds_crop_function)(input, output, output_ptr, input_offset, window_step_x, cols_out_of_bounds[0],
178                                        output->info()->dimension(1) - cols_out_of_bounds[1], input_has_single_channel, is_width_flipped);
179         }
180         // Fill all elements after the in bounds elements with the extrapolation value.
181         if(has_cols_out_of_bounds_after)
182         {
183             out_of_bounds_crop_window(output, output_ptr, extrapolation_value, window_step_x, output->info()->dimension(1) - cols_out_of_bounds[1], output->info()->dimension(1));
184         }
185         output_ptr += output->info()->dimension(1) * output->info()->dimension(0);
186     }
187     // Fill all rows after the in bounds elements with the extrapolation value.
188     out_of_bounds_crop_window(output, output_ptr, extrapolation_value, window_step_x, 0, rows_out_of_bounds[1] * output->info()->dimension(1));
189 }
190 } // namespace
191 
NECropKernel()192 NECropKernel::NECropKernel()
193     : _input(nullptr), _crop_boxes(nullptr), _box_ind(nullptr), _output(nullptr), _start(), _end(), _crop_box_ind(0), _extrapolation_value(0), _rows_out_of_bounds(), _cols_out_of_bounds()
194 {
195 }
196 
configure(const ITensor * input,const ITensor * crop_boxes,const ITensor * box_ind,ITensor * output,uint32_t crop_box_ind,float extrapolation_value)197 void NECropKernel::configure(const ITensor *input, const ITensor *crop_boxes, const ITensor *box_ind, ITensor *output, uint32_t crop_box_ind, float extrapolation_value)
198 {
199     ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
200     ARM_COMPUTE_ERROR_THROW_ON(validate(input->info(), crop_boxes->info(), box_ind->info(), output->info(), crop_box_ind, extrapolation_value));
201 
202     _input               = input;
203     _crop_boxes          = crop_boxes;
204     _box_ind             = box_ind;
205     _output              = output;
206     _crop_box_ind        = crop_box_ind;
207     _extrapolation_value = extrapolation_value;
208 }
209 
validate(const ITensorInfo * input,const ITensorInfo * crop_boxes,const ITensorInfo * box_ind,const ITensorInfo * output,uint32_t crop_box_ind,float extrapolation_value)210 Status NECropKernel::validate(const ITensorInfo *input, const ITensorInfo *crop_boxes, const ITensorInfo *box_ind, const ITensorInfo *output, uint32_t crop_box_ind, float extrapolation_value)
211 {
212     ARM_COMPUTE_UNUSED(extrapolation_value);
213     const auto *uk = get_implementation(CropSelectorData{ input->data_type() });
214     ARM_COMPUTE_RETURN_ERROR_ON(uk == nullptr || uk->ukernel == nullptr);
215 
216     ARM_COMPUTE_RETURN_ERROR_ON_CPU_F16_UNSUPPORTED(input);
217     ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8, DataType::U16, DataType::S16, DataType::F16, DataType::U32, DataType::S32, DataType::F32);
218     ARM_COMPUTE_RETURN_ERROR_ON_DATA_LAYOUT_NOT_IN(input, DataLayout::NHWC);
219     ARM_COMPUTE_RETURN_ERROR_ON(input->tensor_shape().num_dimensions() > 4);
220     ARM_COMPUTE_RETURN_ERROR_ON(crop_boxes->tensor_shape()[0] != 4);
221     ARM_COMPUTE_RETURN_ERROR_ON(crop_boxes->tensor_shape()[1] != box_ind->tensor_shape()[0]);
222     ARM_COMPUTE_RETURN_ERROR_ON(crop_boxes->tensor_shape()[1] <= crop_box_ind);
223     ARM_COMPUTE_RETURN_ERROR_ON(box_ind->tensor_shape()[0] <= crop_box_ind);
224     if(output->total_size() > 0)
225     {
226         ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_NOT_IN(output, DataType::F32);
227         ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_LAYOUT(input, output);
228         ARM_COMPUTE_RETURN_ERROR_ON(output->num_dimensions() != 3);
229         ARM_COMPUTE_RETURN_ERROR_ON(output->has_padding());
230     }
231     return Status{};
232 }
233 
configure_output_shape()234 void NECropKernel::configure_output_shape()
235 {
236     // _crop_box_ind is used to index _crop_boxes and retrieve the appropriate crop box.
237     // The crop box is specified by normalized coordinates [y0, x0, y1, x1].
238     const float x0 = *reinterpret_cast<const float *>(_crop_boxes->ptr_to_element(Coordinates(1, _crop_box_ind)));
239     const float y0 = *reinterpret_cast<const float *>(_crop_boxes->ptr_to_element(Coordinates(0, _crop_box_ind)));
240     const float x1 = *reinterpret_cast<const float *>(_crop_boxes->ptr_to_element(Coordinates(3, _crop_box_ind)));
241     const float y1 = *reinterpret_cast<const float *>(_crop_boxes->ptr_to_element(Coordinates(2, _crop_box_ind)));
242     // The normalized coordiantes are scaled to retrieve the floating point image coordinates which are rounded to integers.
243     _start = Coordinates(std::floor(x0 * (_input->info()->tensor_shape()[1] - 1) + 0.5f),
244                          std::floor(y0 * (_input->info()->tensor_shape()[2] - 1) + 0.5f));
245     _end = Coordinates(std::floor(x1 * (_input->info()->tensor_shape()[1] - 1) + 0.5f),
246                        std::floor(y1 * (_input->info()->tensor_shape()[2] - 1) + 0.5f));
247     const TensorShape out_shape(_input->info()->tensor_shape()[0], abs(_end[0] - _start[0]) + 1, abs(_end[1] - _start[1]) + 1);
248     _output->info()->set_tensor_shape(out_shape);
249 
250     bool is_width_flipped  = _end[0] < _start[0];
251     bool is_height_flipped = _end[1] < _start[1];
252     if(is_height_flipped)
253     {
254         _rows_out_of_bounds[0] = _start[1] >= static_cast<int32_t>(_input->info()->dimension(2)) ? std::min(static_cast<uint32_t>(_start[1] - _input->info()->dimension(2) + 1),
255                                                                                                             static_cast<uint32_t>(_output->info()->dimension(2))) :
256                                  0;
257         _rows_out_of_bounds[1] = _end[1] < 0 ? std::min(static_cast<uint32_t>(-_end[1]),
258                                                         static_cast<uint32_t>(_output->info()->dimension(2))) :
259                                  0;
260     }
261     else
262     {
263         _rows_out_of_bounds[0] = _start[1] < 0 ? std::min(static_cast<uint32_t>(-_start[1]),
264                                                           static_cast<uint32_t>(_output->info()->dimension(2))) :
265                                  0;
266         _rows_out_of_bounds[1] = _end[1] >= static_cast<int32_t>(_input->info()->dimension(2)) ? std::min(static_cast<uint32_t>(_end[1] - _input->info()->dimension(2) + 1),
267                                                                                                           static_cast<uint32_t>(_output->info()->dimension(2))) :
268                                  0;
269     }
270     if(is_width_flipped)
271     {
272         _cols_out_of_bounds[0] = _start[0] >= static_cast<int32_t>(_input->info()->dimension(1)) ? std::min(static_cast<uint32_t>(_start[0] - _input->info()->dimension(1) + 1),
273                                                                                                             static_cast<uint32_t>(_output->info()->dimension(1))) :
274                                  0;
275         _cols_out_of_bounds[1] = _end[0] < 0 ? std::min(static_cast<uint32_t>(-_end[0]),
276                                                         static_cast<uint32_t>(_output->info()->dimension(1))) :
277                                  0;
278     }
279     else
280     {
281         _cols_out_of_bounds[0] = _start[0] < 0 ? std::min(static_cast<uint32_t>(-_start[0]),
282                                                           static_cast<uint32_t>(_output->info()->dimension(1))) :
283                                  0;
284         _cols_out_of_bounds[1] = _end[0] >= static_cast<int32_t>(_input->info()->dimension(1)) ? std::min(static_cast<uint32_t>(_end[0] - _input->info()->dimension(1) + 1),
285                                                                                                           static_cast<uint32_t>(_output->info()->dimension(1))) :
286                                  0;
287     }
288 
289     INEKernel::configure(calculate_max_window(*_output->info()));
290 }
291 
run(const Window & window,const ThreadInfo & info)292 void NECropKernel::run(const Window &window, const ThreadInfo &info)
293 {
294     ARM_COMPUTE_UNUSED(window, info);
295     ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
296     ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(INEKernel::window(), window);
297 
298     ARM_COMPUTE_ERROR_ON(_input->info()->has_padding());
299     ARM_COMPUTE_ERROR_ON(_output->info()->has_padding());
300 
301     const auto *uk = get_implementation(CropSelectorData{ _input->info()->data_type() });
302 
303     uint32_t    batch_index = *(reinterpret_cast<int32_t *>(_box_ind->ptr_to_element(Coordinates(_crop_box_ind))));
304     Coordinates input_offset(0, _end[0] < _start[0] ? _start[0] - _cols_out_of_bounds[0] : _start[0] + _cols_out_of_bounds[0],
305                              _end[1] < _start[1] ? _start[1] - _rows_out_of_bounds[0] : _start[1] + _rows_out_of_bounds[0], batch_index);
306     execute_window(_input, _output, input_offset, _extrapolation_value, _rows_out_of_bounds, _cols_out_of_bounds, uk->ukernel, _end[1] < _start[1],
307                    _cols_out_of_bounds[0] + _cols_out_of_bounds[1] < _output->info()->dimension(1), _cols_out_of_bounds[0] > 0, _cols_out_of_bounds[1] > 0,
308                    _start[0] <= _end[0], _end[0] < _start[0]);
309 }
310 } // namespace arm_compute
311