• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * Copyright (c) 2018-2020 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/CL/kernels/CLBatchToSpaceLayerKernel.h"
25 
26 #include "arm_compute/core/CL/CLHelpers.h"
27 #include "arm_compute/core/CL/ICLTensor.h"
28 #include "arm_compute/core/utils/misc/ShapeCalculator.h"
29 #include "src/core/CL/CLValidate.h"
30 #include "src/core/helpers/AutoConfiguration.h"
31 #include "src/core/helpers/WindowHelpers.h"
32 #include "support/StringSupport.h"
33 
34 using namespace arm_compute::misc::shape_calculator;
35 namespace arm_compute
36 {
37 namespace
38 {
validate_arguments(const ITensorInfo * input,const ITensorInfo * block_info,const ITensorInfo * output)39 Status validate_arguments(const ITensorInfo *input, const ITensorInfo *block_info, const ITensorInfo *output)
40 {
41     ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, block_info, output);
42     ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(block_info, 1, DataType::S32);
43     ARM_COMPUTE_RETURN_ERROR_ON(input->num_dimensions() > 4);
44     ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
45 
46     // Validate output if initialized
47     if(output->total_size() != 0)
48     {
49         ARM_COMPUTE_RETURN_ERROR_ON(output->num_dimensions() > 4);
50         ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
51     }
52 
53     return Status{};
54 }
validate_arguments_static(const ITensorInfo * input,const int block_shape_x,const int block_shape_y,const ITensorInfo * output)55 Status validate_arguments_static(const ITensorInfo *input, const int block_shape_x, const int block_shape_y, const ITensorInfo *output)
56 {
57     ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
58     ARM_COMPUTE_RETURN_ERROR_ON(input->num_dimensions() > 4);
59     ARM_COMPUTE_RETURN_ERROR_ON(block_shape_x <= 0);
60     ARM_COMPUTE_RETURN_ERROR_ON(block_shape_y <= 0);
61 
62     const DataLayout data_layout = input->data_layout();
63     const int        idx_batch   = get_data_layout_dimension_index(data_layout, DataLayoutDimension::BATCHES);
64     ARM_COMPUTE_RETURN_ERROR_ON(input->tensor_shape()[idx_batch] % (block_shape_x * block_shape_y) != 0);
65 
66     // Validate output if initialized
67     if(output->total_size() != 0)
68     {
69         const int idx_width   = get_data_layout_dimension_index(data_layout, DataLayoutDimension::WIDTH);
70         const int idx_height  = get_data_layout_dimension_index(data_layout, DataLayoutDimension::HEIGHT);
71         const int idx_channel = get_data_layout_dimension_index(data_layout, DataLayoutDimension::CHANNEL);
72         ARM_COMPUTE_RETURN_ERROR_ON(output->tensor_shape()[idx_width] != (block_shape_x * input->tensor_shape()[idx_width]));
73         ARM_COMPUTE_RETURN_ERROR_ON(output->tensor_shape()[idx_height] != (block_shape_y * input->tensor_shape()[idx_height]));
74         ARM_COMPUTE_RETURN_ERROR_ON(output->tensor_shape()[idx_channel] != input->tensor_shape()[idx_channel]);
75         ARM_COMPUTE_RETURN_ERROR_ON(output->num_dimensions() > 4);
76         ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
77     }
78 
79     return Status{};
80 }
81 } // namespace
82 
CLBatchToSpaceLayerKernel()83 CLBatchToSpaceLayerKernel::CLBatchToSpaceLayerKernel()
84     : _input(nullptr), _block_shape(nullptr), _output(nullptr)
85 {
86 }
87 
configure(const ICLTensor * input,const ICLTensor * block_shape,ICLTensor * output)88 void CLBatchToSpaceLayerKernel::configure(const ICLTensor *input, const ICLTensor *block_shape, ICLTensor *output)
89 {
90     configure(CLKernelLibrary::get().get_compile_context(), input, block_shape, output);
91 }
92 
configure(const CLCompileContext & compile_context,const ICLTensor * input,const ICLTensor * block_shape,ICLTensor * output)93 void CLBatchToSpaceLayerKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, const ICLTensor *block_shape, ICLTensor *output)
94 {
95     ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
96 
97     auto padding_info = get_padding_info({ input, block_shape, output });
98 
99     ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), block_shape->info(), output->info()));
100 
101     _input       = input;
102     _block_shape = block_shape;
103     _output      = output;
104 
105     const int idx_width = get_data_layout_dimension_index(input->info()->data_layout(), DataLayoutDimension::WIDTH);
106 
107     // Create kernel
108     CLBuildOptions build_opts;
109     build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(input->info()->data_type()));
110     build_opts.add_option("-DBATCH_SIZE=" + support::cpp11::to_string(input->info()->dimension(3)));
111     build_opts.add_option("-DWIDTH_IN=" + support::cpp11::to_string(input->info()->dimension(idx_width)));
112     _kernel = create_kernel(compile_context, "batch_to_space_" + lower_string(string_from_data_layout(input->info()->data_layout())), build_opts.options());
113 
114     // Configure kernel window
115     Window win = calculate_max_window(*input->info(), Steps());
116     ICLKernel::configure_internal(win);
117 
118     ARM_COMPUTE_ERROR_ON(has_padding_changed(padding_info));
119 }
120 
configure(const ICLTensor * input,const int32_t block_shape_x,const int32_t block_shape_y,ICLTensor * output)121 void CLBatchToSpaceLayerKernel::configure(const ICLTensor *input, const int32_t block_shape_x, const int32_t block_shape_y, ICLTensor *output)
122 {
123     configure(CLKernelLibrary::get().get_compile_context(), input, block_shape_x, block_shape_y, output);
124 }
125 
configure(const CLCompileContext & compile_context,const ICLTensor * input,const int32_t block_shape_x,const int32_t block_shape_y,ICLTensor * output)126 void CLBatchToSpaceLayerKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, const int32_t block_shape_x, const int32_t block_shape_y, ICLTensor *output)
127 {
128     ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
129 
130     TensorShape output_shape = compute_batch_to_space_shape(input->info(), block_shape_x, block_shape_y);
131     auto_init_if_empty(*output->info(), output_shape, 1, input->info()->data_type());
132 
133     ARM_COMPUTE_ERROR_THROW_ON(validate_arguments_static(input->info(), block_shape_x, block_shape_y, output->info()));
134 
135     _input  = input;
136     _output = output;
137 
138     const int idx_width = get_data_layout_dimension_index(input->info()->data_layout(), DataLayoutDimension::WIDTH);
139 
140     // Create kernel
141     CLBuildOptions build_opts;
142     build_opts.add_option("-DDATA_TYPE=" + get_cl_unsigned_type_from_element_size(data_size_from_type(input->info()->data_type())));
143     build_opts.add_option("-DBATCH_SIZE=" + support::cpp11::to_string(input->info()->dimension(3)));
144     build_opts.add_option("-DBLOCK_SHAPE_X=" + support::cpp11::to_string(block_shape_x));
145     build_opts.add_option("-DBLOCK_SHAPE_Y=" + support::cpp11::to_string(block_shape_y));
146     build_opts.add_option("-DWIDTH_IN=" + support::cpp11::to_string(input->info()->dimension(idx_width)));
147     _kernel = create_kernel(compile_context, "batch_to_space_static_" + lower_string(string_from_data_layout(input->info()->data_layout())), build_opts.options());
148 
149     // Configure kernel window
150     Window win = calculate_max_window(*input->info(), Steps());
151     ICLKernel::configure_internal(win);
152 }
153 
validate(const ITensorInfo * input,const ITensorInfo * block_shape,const ITensorInfo * output)154 Status CLBatchToSpaceLayerKernel::validate(const ITensorInfo *input, const ITensorInfo *block_shape, const ITensorInfo *output)
155 {
156     ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, block_shape, output);
157     ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, block_shape, output));
158     return Status{};
159 }
160 
validate(const ITensorInfo * input,const int32_t block_shape_x,const int32_t block_shape_y,const ITensorInfo * output)161 Status CLBatchToSpaceLayerKernel::validate(const ITensorInfo *input, const int32_t block_shape_x, const int32_t block_shape_y, const ITensorInfo *output)
162 {
163     ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
164     ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments_static(input, block_shape_x, block_shape_y, output));
165     return Status{};
166 }
167 
run(const Window & window,cl::CommandQueue & queue)168 void CLBatchToSpaceLayerKernel::run(const Window &window, cl::CommandQueue &queue)
169 {
170     ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
171     ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
172 
173     Window slice_in  = window.first_slice_window_3D();
174     Window slice_out = window.first_slice_window_4D();
175 
176     Window vector_slice = window.first_slice_window_1D();
177     vector_slice.set(Window::DimX, Window::Dimension(0, 0, 0));
178 
179     slice_out.set(Window::DimX, Window::Dimension(0, 0, 0));
180     slice_out.set(Window::DimY, Window::Dimension(0, 0, 0));
181     slice_out.set(Window::DimZ, Window::Dimension(0, 0, 0));
182     slice_out.set(3, Window::Dimension(0, 0, 0));
183 
184     int batch_id = 0;
185     do
186     {
187         unsigned int idx = 0;
188         add_3D_tensor_argument(idx, _input, slice_in);
189         add_argument(idx, batch_id);
190         if(_block_shape != nullptr)
191         {
192             add_1D_tensor_argument(idx, _block_shape, vector_slice);
193         }
194         add_4D_tensor_argument(idx, _output, slice_out);
195         enqueue(queue, *this, slice_in, lws_hint());
196 
197         ++batch_id;
198     }
199     while(window.slide_window_slice_3D(slice_in));
200 }
201 } // namespace arm_compute
202