cxx11_tensor_shuffling.cpp (7692B)
1 // This file is part of Eigen, a lightweight C++ template library 2 // for linear algebra. 3 // 4 // Copyright (C) 2014 Benoit Steiner <benoit.steiner.goog@gmail.com> 5 // 6 // This Source Code Form is subject to the terms of the Mozilla 7 // Public License v. 2.0. If a copy of the MPL was not distributed 8 // with this file, You can obtain one at http://mozilla.org/MPL/2.0/. 9 10 #include "main.h" 11 12 #include <Eigen/CXX11/Tensor> 13 14 using Eigen::Tensor; 15 using Eigen::array; 16 17 template <int DataLayout> 18 static void test_simple_shuffling() 19 { 20 Tensor<float, 4, DataLayout> tensor(2,3,5,7); 21 tensor.setRandom(); 22 array<ptrdiff_t, 4> shuffles; 23 shuffles[0] = 0; 24 shuffles[1] = 1; 25 shuffles[2] = 2; 26 shuffles[3] = 3; 27 28 Tensor<float, 4, DataLayout> no_shuffle; 29 no_shuffle = tensor.shuffle(shuffles); 30 31 VERIFY_IS_EQUAL(no_shuffle.dimension(0), 2); 32 VERIFY_IS_EQUAL(no_shuffle.dimension(1), 3); 33 VERIFY_IS_EQUAL(no_shuffle.dimension(2), 5); 34 VERIFY_IS_EQUAL(no_shuffle.dimension(3), 7); 35 36 for (int i = 0; i < 2; ++i) { 37 for (int j = 0; j < 3; ++j) { 38 for (int k = 0; k < 5; ++k) { 39 for (int l = 0; l < 7; ++l) { 40 VERIFY_IS_EQUAL(tensor(i,j,k,l), no_shuffle(i,j,k,l)); 41 } 42 } 43 } 44 } 45 46 shuffles[0] = 2; 47 shuffles[1] = 3; 48 shuffles[2] = 1; 49 shuffles[3] = 0; 50 Tensor<float, 4, DataLayout> shuffle; 51 shuffle = tensor.shuffle(shuffles); 52 53 VERIFY_IS_EQUAL(shuffle.dimension(0), 5); 54 VERIFY_IS_EQUAL(shuffle.dimension(1), 7); 55 VERIFY_IS_EQUAL(shuffle.dimension(2), 3); 56 VERIFY_IS_EQUAL(shuffle.dimension(3), 2); 57 58 for (int i = 0; i < 2; ++i) { 59 for (int j = 0; j < 3; ++j) { 60 for (int k = 0; k < 5; ++k) { 61 for (int l = 0; l < 7; ++l) { 62 VERIFY_IS_EQUAL(tensor(i,j,k,l), shuffle(k,l,j,i)); 63 } 64 } 65 } 66 } 67 } 68 69 70 template <int DataLayout> 71 static void test_expr_shuffling() 72 { 73 Tensor<float, 4, DataLayout> tensor(2,3,5,7); 74 tensor.setRandom(); 75 76 array<ptrdiff_t, 4> shuffles; 77 shuffles[0] = 2; 78 shuffles[1] = 3; 79 shuffles[2] = 1; 80 shuffles[3] = 0; 81 Tensor<float, 4, DataLayout> expected; 82 expected = tensor.shuffle(shuffles); 83 84 Tensor<float, 4, DataLayout> result(5, 7, 3, 2); 85 86 array<ptrdiff_t, 4> src_slice_dim{{2, 3, 1, 7}}; 87 array<ptrdiff_t, 4> src_slice_start{{0, 0, 0, 0}}; 88 array<ptrdiff_t, 4> dst_slice_dim{{1, 7, 3, 2}}; 89 array<ptrdiff_t, 4> dst_slice_start{{0, 0, 0, 0}}; 90 91 for (int i = 0; i < 5; ++i) { 92 result.slice(dst_slice_start, dst_slice_dim) = 93 tensor.slice(src_slice_start, src_slice_dim).shuffle(shuffles); 94 src_slice_start[2] += 1; 95 dst_slice_start[0] += 1; 96 } 97 98 VERIFY_IS_EQUAL(result.dimension(0), 5); 99 VERIFY_IS_EQUAL(result.dimension(1), 7); 100 VERIFY_IS_EQUAL(result.dimension(2), 3); 101 VERIFY_IS_EQUAL(result.dimension(3), 2); 102 103 for (int i = 0; i < expected.dimension(0); ++i) { 104 for (int j = 0; j < expected.dimension(1); ++j) { 105 for (int k = 0; k < expected.dimension(2); ++k) { 106 for (int l = 0; l < expected.dimension(3); ++l) { 107 VERIFY_IS_EQUAL(result(i,j,k,l), expected(i,j,k,l)); 108 } 109 } 110 } 111 } 112 113 dst_slice_start[0] = 0; 114 result.setRandom(); 115 for (int i = 0; i < 5; ++i) { 116 result.slice(dst_slice_start, dst_slice_dim) = 117 tensor.shuffle(shuffles).slice(dst_slice_start, dst_slice_dim); 118 dst_slice_start[0] += 1; 119 } 120 121 for (int i = 0; i < expected.dimension(0); ++i) { 122 for (int j = 0; j < expected.dimension(1); ++j) { 123 for (int k = 0; k < expected.dimension(2); ++k) { 124 for (int l = 0; l < expected.dimension(3); ++l) { 125 VERIFY_IS_EQUAL(result(i,j,k,l), expected(i,j,k,l)); 126 } 127 } 128 } 129 } 130 } 131 132 133 template <int DataLayout> 134 static void test_shuffling_as_value() 135 { 136 Tensor<float, 4, DataLayout> tensor(2,3,5,7); 137 tensor.setRandom(); 138 array<ptrdiff_t, 4> shuffles; 139 shuffles[2] = 0; 140 shuffles[3] = 1; 141 shuffles[1] = 2; 142 shuffles[0] = 3; 143 Tensor<float, 4, DataLayout> shuffle(5,7,3,2); 144 shuffle.shuffle(shuffles) = tensor; 145 146 VERIFY_IS_EQUAL(shuffle.dimension(0), 5); 147 VERIFY_IS_EQUAL(shuffle.dimension(1), 7); 148 VERIFY_IS_EQUAL(shuffle.dimension(2), 3); 149 VERIFY_IS_EQUAL(shuffle.dimension(3), 2); 150 151 for (int i = 0; i < 2; ++i) { 152 for (int j = 0; j < 3; ++j) { 153 for (int k = 0; k < 5; ++k) { 154 for (int l = 0; l < 7; ++l) { 155 VERIFY_IS_EQUAL(tensor(i,j,k,l), shuffle(k,l,j,i)); 156 } 157 } 158 } 159 } 160 161 array<ptrdiff_t, 4> no_shuffle; 162 no_shuffle[0] = 0; 163 no_shuffle[1] = 1; 164 no_shuffle[2] = 2; 165 no_shuffle[3] = 3; 166 Tensor<float, 4, DataLayout> shuffle2(5,7,3,2); 167 shuffle2.shuffle(shuffles) = tensor.shuffle(no_shuffle); 168 for (int i = 0; i < 5; ++i) { 169 for (int j = 0; j < 7; ++j) { 170 for (int k = 0; k < 3; ++k) { 171 for (int l = 0; l < 2; ++l) { 172 VERIFY_IS_EQUAL(shuffle2(i,j,k,l), shuffle(i,j,k,l)); 173 } 174 } 175 } 176 } 177 } 178 179 180 template <int DataLayout> 181 static void test_shuffle_unshuffle() 182 { 183 Tensor<float, 4, DataLayout> tensor(2,3,5,7); 184 tensor.setRandom(); 185 186 // Choose a random permutation. 187 array<ptrdiff_t, 4> shuffles; 188 for (int i = 0; i < 4; ++i) { 189 shuffles[i] = i; 190 } 191 array<ptrdiff_t, 4> shuffles_inverse; 192 for (int i = 0; i < 4; ++i) { 193 const ptrdiff_t index = internal::random<ptrdiff_t>(i, 3); 194 shuffles_inverse[shuffles[index]] = i; 195 std::swap(shuffles[i], shuffles[index]); 196 } 197 198 Tensor<float, 4, DataLayout> shuffle; 199 shuffle = tensor.shuffle(shuffles).shuffle(shuffles_inverse); 200 201 VERIFY_IS_EQUAL(shuffle.dimension(0), 2); 202 VERIFY_IS_EQUAL(shuffle.dimension(1), 3); 203 VERIFY_IS_EQUAL(shuffle.dimension(2), 5); 204 VERIFY_IS_EQUAL(shuffle.dimension(3), 7); 205 206 for (int i = 0; i < 2; ++i) { 207 for (int j = 0; j < 3; ++j) { 208 for (int k = 0; k < 5; ++k) { 209 for (int l = 0; l < 7; ++l) { 210 VERIFY_IS_EQUAL(tensor(i,j,k,l), shuffle(i,j,k,l)); 211 } 212 } 213 } 214 } 215 } 216 217 218 template <int DataLayout> 219 static void test_empty_shuffling() 220 { 221 Tensor<float, 4, DataLayout> tensor(2,3,0,7); 222 tensor.setRandom(); 223 array<ptrdiff_t, 4> shuffles; 224 shuffles[0] = 0; 225 shuffles[1] = 1; 226 shuffles[2] = 2; 227 shuffles[3] = 3; 228 229 Tensor<float, 4, DataLayout> no_shuffle; 230 no_shuffle = tensor.shuffle(shuffles); 231 232 VERIFY_IS_EQUAL(no_shuffle.dimension(0), 2); 233 VERIFY_IS_EQUAL(no_shuffle.dimension(1), 3); 234 VERIFY_IS_EQUAL(no_shuffle.dimension(2), 0); 235 VERIFY_IS_EQUAL(no_shuffle.dimension(3), 7); 236 237 for (int i = 0; i < 2; ++i) { 238 for (int j = 0; j < 3; ++j) { 239 for (int k = 0; k < 0; ++k) { 240 for (int l = 0; l < 7; ++l) { 241 VERIFY_IS_EQUAL(tensor(i,j,k,l), no_shuffle(i,j,k,l)); 242 } 243 } 244 } 245 } 246 247 shuffles[0] = 2; 248 shuffles[1] = 3; 249 shuffles[2] = 1; 250 shuffles[3] = 0; 251 Tensor<float, 4, DataLayout> shuffle; 252 shuffle = tensor.shuffle(shuffles); 253 254 VERIFY_IS_EQUAL(shuffle.dimension(0), 0); 255 VERIFY_IS_EQUAL(shuffle.dimension(1), 7); 256 VERIFY_IS_EQUAL(shuffle.dimension(2), 3); 257 VERIFY_IS_EQUAL(shuffle.dimension(3), 2); 258 259 for (int i = 0; i < 2; ++i) { 260 for (int j = 0; j < 3; ++j) { 261 for (int k = 0; k < 0; ++k) { 262 for (int l = 0; l < 7; ++l) { 263 VERIFY_IS_EQUAL(tensor(i,j,k,l), shuffle(k,l,j,i)); 264 } 265 } 266 } 267 } 268 } 269 270 271 EIGEN_DECLARE_TEST(cxx11_tensor_shuffling) 272 { 273 CALL_SUBTEST(test_simple_shuffling<ColMajor>()); 274 CALL_SUBTEST(test_simple_shuffling<RowMajor>()); 275 CALL_SUBTEST(test_expr_shuffling<ColMajor>()); 276 CALL_SUBTEST(test_expr_shuffling<RowMajor>()); 277 CALL_SUBTEST(test_shuffling_as_value<ColMajor>()); 278 CALL_SUBTEST(test_shuffling_as_value<RowMajor>()); 279 CALL_SUBTEST(test_shuffle_unshuffle<ColMajor>()); 280 CALL_SUBTEST(test_shuffle_unshuffle<RowMajor>()); 281 CALL_SUBTEST(test_empty_shuffling<ColMajor>()); 282 CALL_SUBTEST(test_empty_shuffling<RowMajor>()); 283 }