1 /*
2 * Copyright (c) 2018-2021 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/CLROIAlignLayerKernel.h"
25
26 #include "arm_compute/core/CL/CLHelpers.h"
27 #include "arm_compute/core/CL/CLKernelLibrary.h"
28 #include "arm_compute/core/CL/ICLTensor.h"
29 #include "arm_compute/core/CL/OpenCL.h"
30 #include "arm_compute/core/Helpers.h"
31 #include "arm_compute/core/TensorInfo.h"
32 #include "arm_compute/core/Utils.h"
33 #include "arm_compute/core/utils/misc/ShapeCalculator.h"
34 #include "src/core/CL/CLValidate.h"
35 #include "src/core/helpers/AutoConfiguration.h"
36 #include "src/core/helpers/WindowHelpers.h"
37 #include "support/StringSupport.h"
38
39 using namespace arm_compute::misc::shape_calculator;
40
41 namespace arm_compute
42 {
43 namespace
44 {
validate_arguments(const ITensorInfo * input,const ITensorInfo * rois,ITensorInfo * output,const ROIPoolingLayerInfo & pool_info)45 Status validate_arguments(const ITensorInfo *input, const ITensorInfo *rois, ITensorInfo *output, const ROIPoolingLayerInfo &pool_info)
46 {
47 ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, rois, output);
48 ARM_COMPUTE_RETURN_ERROR_ON(rois->dimension(0) != 5);
49 ARM_COMPUTE_RETURN_ERROR_ON(rois->num_dimensions() > 2);
50 ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(input);
51 ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::QASYMM8, DataType::QASYMM8_SIGNED, DataType::F32, DataType::F16);
52 ARM_COMPUTE_RETURN_ERROR_ON_DATA_LAYOUT_NOT_IN(input, DataLayout::NHWC, DataLayout::NCHW);
53 ARM_COMPUTE_RETURN_ERROR_ON((pool_info.pooled_width() == 0) || (pool_info.pooled_height() == 0));
54
55 if(output->total_size() != 0)
56 {
57 ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
58 ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_LAYOUT(input, output);
59 ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DIMENSIONS(compute_roi_align_shape(*input, *rois, pool_info), output->tensor_shape());
60 }
61
62 if(is_data_type_quantized_asymmetric(input->data_type()))
63 {
64 ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(rois, 1, DataType::QASYMM16);
65
66 const UniformQuantizationInfo rois_qinfo = rois->quantization_info().uniform();
67 ARM_COMPUTE_RETURN_ERROR_ON(rois_qinfo.scale != 0.125f);
68 ARM_COMPUTE_RETURN_ERROR_ON(rois_qinfo.offset != 0);
69 }
70 else
71 {
72 ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, rois);
73 }
74 return Status{};
75 }
76
77 } // namespace
78
CLROIAlignLayerKernel()79 CLROIAlignLayerKernel::CLROIAlignLayerKernel()
80 : _input(nullptr), _output(nullptr), _rois(nullptr), _pool_info(0, 0, 0.f)
81 {
82 _type = CLKernelType::ELEMENTWISE;
83 }
84
configure(const ICLTensor * input,const ICLTensor * rois,ICLTensor * output,const ROIPoolingLayerInfo & pool_info)85 void CLROIAlignLayerKernel::configure(const ICLTensor *input, const ICLTensor *rois, ICLTensor *output, const ROIPoolingLayerInfo &pool_info)
86 {
87 configure(CLKernelLibrary::get().get_compile_context(), input, rois, output, pool_info);
88 }
89
configure(const CLCompileContext & compile_context,const ICLTensor * input,const ICLTensor * rois,ICLTensor * output,const ROIPoolingLayerInfo & pool_info)90 void CLROIAlignLayerKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, const ICLTensor *rois, ICLTensor *output, const ROIPoolingLayerInfo &pool_info)
91 {
92 ARM_COMPUTE_ERROR_ON_NULLPTR(input, output, rois);
93 ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), rois->info(), output->info(), pool_info));
94
95 // Output auto inizialitation if not yet initialized
96 const TensorShape output_shape = compute_roi_align_shape(*input->info(), *rois->info(), pool_info);
97 auto_init_if_empty(*output->info(), output_shape, 1, input->info()->data_type());
98 output->info()->set_data_layout(input->info()->data_layout());
99
100 auto padding_info = get_padding_info({ input, rois, output });
101
102 _input = input;
103 _output = output;
104 _rois = rois;
105 _pool_info = pool_info;
106
107 const DataType data_type = input->info()->data_type();
108 const bool is_qasymm = is_data_type_quantized_asymmetric(data_type);
109
110 // Set build options
111 CLBuildOptions build_opts;
112 build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(data_type));
113 build_opts.add_option("-DDATA_SIZE=" + get_data_size_from_data_type(input->info()->data_type()));
114 build_opts.add_option("-DMAX_DIM_X=" + support::cpp11::to_string(_input->info()->dimension(get_data_layout_dimension_index(input->info()->data_layout(), DataLayoutDimension::WIDTH))));
115 build_opts.add_option("-DMAX_DIM_Y=" + support::cpp11::to_string(_input->info()->dimension(get_data_layout_dimension_index(input->info()->data_layout(), DataLayoutDimension::HEIGHT))));
116 build_opts.add_option("-DMAX_DIM_Z=" + support::cpp11::to_string(_input->info()->dimension(get_data_layout_dimension_index(input->info()->data_layout(), DataLayoutDimension::CHANNEL))));
117 build_opts.add_option("-DPOOLED_DIM_X=" + support::cpp11::to_string(pool_info.pooled_width()));
118 build_opts.add_option("-DPOOLED_DIM_Y=" + support::cpp11::to_string(pool_info.pooled_height()));
119 build_opts.add_option("-DSPATIAL_SCALE=" + float_to_string_with_full_precision(pool_info.spatial_scale()));
120 build_opts.add_option_if(input->info()->data_layout() == DataLayout::NHWC, "-DNHWC");
121 build_opts.add_option_if(pool_info.sampling_ratio() > 0, "-DSAMPLING_RATIO=" + support::cpp11::to_string(pool_info.sampling_ratio()));
122
123 if(is_qasymm)
124 {
125 const UniformQuantizationInfo iq_info = input->info()->quantization_info().uniform();
126 const UniformQuantizationInfo roisq_info = rois->info()->quantization_info().uniform();
127 const UniformQuantizationInfo oq_info = output->info()->quantization_info().uniform();
128
129 build_opts.add_option("-DOFFSET_IN=" + float_to_string_with_full_precision(iq_info.offset));
130 build_opts.add_option("-DSCALE_IN=" + float_to_string_with_full_precision(iq_info.scale));
131 build_opts.add_option("-DOFFSET_ROIS=" + float_to_string_with_full_precision(roisq_info.offset));
132 build_opts.add_option("-DSCALE_ROIS=" + float_to_string_with_full_precision(roisq_info.scale));
133 build_opts.add_option("-DOFFSET_OUT=" + float_to_string_with_full_precision(oq_info.offset));
134 build_opts.add_option("-DSCALE_OUT=" + float_to_string_with_full_precision(oq_info.scale));
135 }
136
137 // Create kernel
138 const std::string kernel_name = (is_qasymm) ? "roi_align_layer_quantized" : "roi_align_layer";
139 _kernel = create_kernel(compile_context, kernel_name, build_opts.options());
140
141 // Configure kernel window
142 Window win = calculate_max_window(*output->info(), Steps());
143 ICLKernel::configure_internal(win);
144 ARM_COMPUTE_ERROR_ON(has_padding_changed(padding_info));
145 }
146
validate(const ITensorInfo * input,const ITensorInfo * rois,ITensorInfo * output,const ROIPoolingLayerInfo & pool_info)147 Status CLROIAlignLayerKernel::validate(const ITensorInfo *input, const ITensorInfo *rois, ITensorInfo *output, const ROIPoolingLayerInfo &pool_info)
148 {
149 ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, rois, output, pool_info));
150 return Status{};
151 }
152
run(const Window & window,cl::CommandQueue & queue)153 void CLROIAlignLayerKernel::run(const Window &window, cl::CommandQueue &queue)
154 {
155 ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
156 ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(IKernel::window(), window);
157
158 Window slice = window.first_slice_window_3D();
159 Window slice_rois = slice;
160 // Parallelize spatially and across the fourth dimension of the output tensor (also across ROITensor)
161 slice_rois.set_dimension_step(Window::DimX, _rois->info()->dimension(0));
162 slice.set(get_data_layout_dimension_index(_input->info()->data_layout(), DataLayoutDimension::CHANNEL), window[3]);
163
164 // Set arguments
165 unsigned int idx = 0;
166 add_3D_tensor_argument(idx, _input, slice);
167 add_2D_tensor_argument(idx, _rois, slice_rois);
168 add_3D_tensor_argument(idx, _output, slice);
169 add_argument<cl_uint>(idx, _input->info()->strides_in_bytes()[3]);
170 add_argument<cl_uint>(idx, _output->info()->strides_in_bytes()[3]);
171
172 enqueue(queue, *this, slice, lws_hint());
173 }
174 } // namespace arm_compute
175