Marat Dukhan | 4662b19 | 2020-05-21 15:52:03 -0700 | [diff] [blame] | 1 | // Copyright 2019 Google LLC |
| 2 | // |
| 3 | // This source code is licensed under the BSD-style license found in the |
| 4 | // LICENSE file in the root directory of this source tree. |
| 5 | |
| 6 | #pragma once |
| 7 | |
| 8 | #include <gtest/gtest.h> |
| 9 | |
| 10 | #include <algorithm> |
| 11 | #include <array> |
| 12 | #include <cstddef> |
| 13 | #include <cstdlib> |
| 14 | #include <functional> |
| 15 | #include <initializer_list> |
| 16 | #include <numeric> |
| 17 | #include <random> |
| 18 | #include <vector> |
| 19 | |
| 20 | #include <xnnpack.h> |
| 21 | |
| 22 | |
Marat Dukhan | 065b11e | 2020-05-22 09:49:41 -0700 | [diff] [blame] | 23 | class ConstantPadOperatorTester { |
Marat Dukhan | 4662b19 | 2020-05-21 15:52:03 -0700 | [diff] [blame] | 24 | public: |
Marat Dukhan | 065b11e | 2020-05-22 09:49:41 -0700 | [diff] [blame] | 25 | inline ConstantPadOperatorTester& input_shape(std::initializer_list<size_t> input_shape) { |
Marat Dukhan | 4662b19 | 2020-05-21 15:52:03 -0700 | [diff] [blame] | 26 | assert(input_shape.size() <= XNN_MAX_TENSOR_DIMS); |
| 27 | input_shape_ = std::vector<size_t>(input_shape); |
| 28 | return *this; |
| 29 | } |
| 30 | |
| 31 | inline const std::vector<size_t>& input_shape() const { |
| 32 | return input_shape_; |
| 33 | } |
| 34 | |
| 35 | inline size_t input_dim(size_t i) const { |
| 36 | return i < input_shape_.size() ? input_shape_[i] : 1; |
| 37 | } |
| 38 | |
| 39 | inline size_t num_dims() const { |
| 40 | return input_shape_.size(); |
| 41 | } |
| 42 | |
| 43 | inline size_t num_input_elements() const { |
| 44 | return std::accumulate( |
| 45 | input_shape_.cbegin(), input_shape_.cend(), size_t(1), std::multiplies<size_t>()); |
| 46 | } |
| 47 | |
Marat Dukhan | 065b11e | 2020-05-22 09:49:41 -0700 | [diff] [blame] | 48 | inline ConstantPadOperatorTester& pre_paddings(std::initializer_list<size_t> pre_paddings) { |
Marat Dukhan | 4662b19 | 2020-05-21 15:52:03 -0700 | [diff] [blame] | 49 | assert(pre_paddings.size() <= XNN_MAX_TENSOR_DIMS); |
| 50 | pre_paddings_ = std::vector<size_t>(pre_paddings); |
| 51 | return *this; |
| 52 | } |
| 53 | |
| 54 | inline const std::vector<size_t>& pre_paddings() const { |
| 55 | return pre_paddings_; |
| 56 | } |
| 57 | |
| 58 | inline size_t pre_padding(size_t i) const { |
| 59 | return i < pre_paddings_.size() ? pre_paddings_[i] : 0; |
| 60 | } |
| 61 | |
| 62 | inline size_t num_pre_paddings() const { |
| 63 | return pre_paddings_.size(); |
| 64 | } |
| 65 | |
Marat Dukhan | 065b11e | 2020-05-22 09:49:41 -0700 | [diff] [blame] | 66 | inline ConstantPadOperatorTester& post_paddings(std::initializer_list<size_t> post_paddings) { |
Marat Dukhan | 4662b19 | 2020-05-21 15:52:03 -0700 | [diff] [blame] | 67 | assert(post_paddings.size() <= XNN_MAX_TENSOR_DIMS); |
| 68 | post_paddings_ = std::vector<size_t>(post_paddings); |
| 69 | return *this; |
| 70 | } |
| 71 | |
| 72 | inline const std::vector<size_t>& post_paddings() const { |
| 73 | return post_paddings_; |
| 74 | } |
| 75 | |
| 76 | inline size_t post_padding(size_t i) const { |
| 77 | return i < post_paddings_.size() ? post_paddings_[i] : 0; |
| 78 | } |
| 79 | |
| 80 | inline size_t num_post_paddings() const { |
| 81 | return post_paddings_.size(); |
| 82 | } |
| 83 | |
| 84 | inline size_t output_dim(size_t i) const { |
| 85 | return pre_padding(i) + input_dim(i) + post_padding(i); |
| 86 | } |
| 87 | |
| 88 | inline size_t num_output_elements() const { |
| 89 | size_t elements = 1; |
| 90 | for (size_t i = 0; i < num_dims(); i++) { |
| 91 | elements *= output_dim(i); |
| 92 | } |
| 93 | return elements; |
| 94 | } |
| 95 | |
Marat Dukhan | 065b11e | 2020-05-22 09:49:41 -0700 | [diff] [blame] | 96 | inline ConstantPadOperatorTester& iterations(size_t iterations) { |
Marat Dukhan | 4662b19 | 2020-05-21 15:52:03 -0700 | [diff] [blame] | 97 | this->iterations_ = iterations; |
| 98 | return *this; |
| 99 | } |
| 100 | |
| 101 | inline size_t iterations() const { |
| 102 | return this->iterations_; |
| 103 | } |
| 104 | |
| 105 | void TestX32() const { |
| 106 | ASSERT_EQ(num_dims(), num_pre_paddings()); |
| 107 | ASSERT_EQ(num_dims(), num_post_paddings()); |
| 108 | |
| 109 | std::random_device random_device; |
| 110 | auto rng = std::mt19937(random_device()); |
| 111 | auto u32rng = std::bind(std::uniform_int_distribution<uint32_t>(), rng); |
| 112 | |
| 113 | // Compute generalized shapes. |
| 114 | std::array<size_t, XNN_MAX_TENSOR_DIMS> input_dims; |
| 115 | std::array<size_t, XNN_MAX_TENSOR_DIMS> input_pre_paddings; |
| 116 | std::array<size_t, XNN_MAX_TENSOR_DIMS> input_post_paddings; |
| 117 | std::array<size_t, XNN_MAX_TENSOR_DIMS> output_dims; |
| 118 | std::fill(input_dims.begin(), input_dims.end(), 1); |
| 119 | std::fill(input_pre_paddings.begin(), input_pre_paddings.end(), 0); |
| 120 | std::fill(input_post_paddings.begin(), input_post_paddings.end(), 0); |
| 121 | std::fill(output_dims.begin(), output_dims.end(), 1); |
| 122 | for (size_t i = 0; i < num_dims(); i++) { |
| 123 | input_dims[XNN_MAX_TENSOR_DIMS - num_dims() + i] = input_dim(i); |
| 124 | input_pre_paddings[XNN_MAX_TENSOR_DIMS - num_dims() + i] = pre_padding(i); |
| 125 | input_post_paddings[XNN_MAX_TENSOR_DIMS - num_dims() + i] = post_padding(i); |
| 126 | output_dims[XNN_MAX_TENSOR_DIMS - num_dims() + i] = output_dim(i); |
| 127 | } |
| 128 | |
| 129 | // Compute generalized strides. |
| 130 | std::array<size_t, XNN_MAX_TENSOR_DIMS> input_strides; |
| 131 | std::array<size_t, XNN_MAX_TENSOR_DIMS> output_strides; |
| 132 | size_t input_stride = 1, output_stride = 1; |
| 133 | for (size_t i = XNN_MAX_TENSOR_DIMS; i != 0; i--) { |
| 134 | input_strides[i - 1] = input_stride; |
| 135 | output_strides[i - 1] = output_stride; |
| 136 | input_stride *= input_dims[i - 1]; |
| 137 | output_stride *= output_dims[i - 1]; |
| 138 | } |
| 139 | |
| 140 | std::vector<uint32_t> input(XNN_EXTRA_BYTES / sizeof(uint32_t) + num_input_elements()); |
| 141 | std::vector<uint32_t> output(num_output_elements()); |
| 142 | std::vector<uint32_t> output_ref(num_output_elements()); |
| 143 | for (size_t iteration = 0; iteration < iterations(); iteration++) { |
| 144 | std::generate(input.begin(), input.end(), std::ref(u32rng)); |
| 145 | std::fill(output.begin(), output.end(), UINT32_C(0xDEADBEEF)); |
| 146 | const uint32_t padding_value = u32rng(); |
| 147 | |
| 148 | // Compute reference results. |
| 149 | std::fill(output_ref.begin(), output_ref.end(), padding_value); |
| 150 | for (size_t i = 0; i < input_dims[0]; i++) { |
| 151 | for (size_t j = 0; j < input_dims[1]; j++) { |
| 152 | for (size_t k = 0; k < input_dims[2]; k++) { |
| 153 | for (size_t l = 0; l < input_dims[3]; l++) { |
| 154 | for (size_t m = 0; m < input_dims[4]; m++) { |
| 155 | for (size_t n = 0; n < input_dims[5]; n++) { |
| 156 | const size_t output_index = |
| 157 | (i + input_pre_paddings[0]) * output_strides[0] + |
| 158 | (j + input_pre_paddings[1]) * output_strides[1] + |
| 159 | (k + input_pre_paddings[2]) * output_strides[2] + |
| 160 | (l + input_pre_paddings[3]) * output_strides[3] + |
| 161 | (m + input_pre_paddings[4]) * output_strides[4] + |
| 162 | (n + input_pre_paddings[5]) * output_strides[5]; |
| 163 | const size_t input_index = |
| 164 | i * input_strides[0] + j * input_strides[1] + k * input_strides[2] + |
| 165 | l * input_strides[3] + m * input_strides[4] + n * input_strides[5]; |
| 166 | output_ref[output_index] = input[input_index]; |
| 167 | } |
| 168 | } |
| 169 | } |
| 170 | } |
| 171 | } |
| 172 | } |
| 173 | |
| 174 | // Create, setup, run, and destroy a binary elementwise operator. |
| 175 | ASSERT_EQ(xnn_status_success, xnn_initialize(nullptr /* allocator */)); |
| 176 | xnn_operator_t pad_op = nullptr; |
| 177 | |
| 178 | ASSERT_EQ(xnn_status_success, |
Marat Dukhan | 065b11e | 2020-05-22 09:49:41 -0700 | [diff] [blame] | 179 | xnn_create_constant_pad_nd_x32( |
Marat Dukhan | 4662b19 | 2020-05-21 15:52:03 -0700 | [diff] [blame] | 180 | &padding_value, 0, &pad_op)); |
| 181 | ASSERT_NE(nullptr, pad_op); |
| 182 | |
| 183 | // Smart pointer to automatically delete pad_op. |
| 184 | std::unique_ptr<xnn_operator, decltype(&xnn_delete_operator)> auto_pad_op(pad_op, xnn_delete_operator); |
| 185 | |
| 186 | ASSERT_EQ(xnn_status_success, |
Marat Dukhan | 065b11e | 2020-05-22 09:49:41 -0700 | [diff] [blame] | 187 | xnn_setup_constant_pad_nd_x32( |
Marat Dukhan | 4662b19 | 2020-05-21 15:52:03 -0700 | [diff] [blame] | 188 | pad_op, |
| 189 | num_dims(), |
| 190 | input_shape().data(), pre_paddings().data(), post_paddings().data(), |
| 191 | input.data(), output.data(), |
| 192 | nullptr /* thread pool */)); |
| 193 | |
| 194 | ASSERT_EQ(xnn_status_success, |
| 195 | xnn_run_operator(pad_op, nullptr /* thread pool */)); |
| 196 | |
| 197 | // Verify results. |
| 198 | for (size_t i = 0; i < output_dims[0]; i++) { |
| 199 | for (size_t j = 0; j < output_dims[1]; j++) { |
| 200 | for (size_t k = 0; k < output_dims[2]; k++) { |
| 201 | for (size_t l = 0; l < output_dims[3]; l++) { |
| 202 | for (size_t m = 0; m < output_dims[4]; m++) { |
| 203 | for (size_t n = 0; n < output_dims[5]; n++) { |
| 204 | const size_t index = |
| 205 | i * output_strides[0] + j * output_strides[1] + k * output_strides[2] + |
| 206 | l * output_strides[3] + m * output_strides[4] + n * output_strides[5]; |
| 207 | ASSERT_EQ(output[index], output_ref[index]) |
| 208 | << "(i, j, k, l, m, n) = (" |
| 209 | << i << ", " << j << ", " << k << ", " << l << ", " << m << ", " << n << ")" |
| 210 | << ", padding value = " << padding_value; |
| 211 | } |
| 212 | } |
| 213 | } |
| 214 | } |
| 215 | } |
| 216 | } |
| 217 | } |
| 218 | } |
| 219 | |
| 220 | private: |
| 221 | std::vector<size_t> input_shape_; |
| 222 | std::vector<size_t> pre_paddings_; |
| 223 | std::vector<size_t> post_paddings_; |
| 224 | size_t iterations_{3}; |
| 225 | }; |