| Line | Branch | Exec | Source |
|---|---|---|---|
| 1 | #include "kapanova_s_image_smoothing/mpi/include/ops_mpi.hpp" | ||
| 2 | |||
| 3 | #include <mpi.h> | ||
| 4 | |||
| 5 | #include <algorithm> | ||
| 6 | #include <array> | ||
| 7 | #include <cmath> | ||
| 8 | #include <cstddef> | ||
| 9 | #include <cstdint> | ||
| 10 | #include <vector> | ||
| 11 | |||
| 12 | #include "kapanova_s_image_smoothing/common/include/common.hpp" | ||
| 13 | |||
| 14 | namespace kapanova_s_image_smoothing { | ||
| 15 | |||
| 16 | namespace { | ||
| 17 | constexpr int kTagExit = 0; | ||
| 18 | constexpr int kTagData = 2; | ||
| 19 | constexpr int kTagResult = 3; | ||
| 20 | constexpr int kRadius = 1; | ||
| 21 | constexpr int kKernelSize = (2 * kRadius) + 1; | ||
| 22 | constexpr float kSigma = 1.5F; | ||
| 23 | constexpr float kSigmaSquared = kSigma * kSigma; | ||
| 24 | constexpr int kEscapeSignal = 0; | ||
| 25 | constexpr int kNoEscapeSignal = 1; | ||
| 26 | } // namespace | ||
| 27 | |||
| 28 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 6 times.
|
6 | KapanovaSImageSmoothingMPI::KapanovaSImageSmoothingMPI(const InType &in) : radius_(kRadius) { |
| 29 | SetTypeOfTask(GetStaticTypeOfTask()); | ||
| 30 |
3/8✗ Branch 0 not taken.
✓ Branch 1 taken 6 times.
✓ Branch 3 taken 6 times.
✗ Branch 4 not taken.
✓ Branch 6 taken 6 times.
✗ Branch 7 not taken.
✗ Branch 10 not taken.
✗ Branch 11 not taken.
|
6 | GetInput() = in.empty() ? InType() : in; |
| 31 | 6 | } | |
| 32 | |||
| 33 | 6 | bool KapanovaSImageSmoothingMPI::ValidationImpl() { | |
| 34 | 6 | int rank = 0; | |
| 35 | 6 | MPI_Comm_rank(MPI_COMM_WORLD, &rank); | |
| 36 | |||
| 37 |
2/2✓ Branch 0 taken 3 times.
✓ Branch 1 taken 3 times.
|
6 | if (rank == 0) { |
| 38 | const auto &input_data = GetInput(); | ||
| 39 |
2/4✓ Branch 0 taken 3 times.
✗ Branch 1 not taken.
✓ Branch 2 taken 3 times.
✗ Branch 3 not taken.
|
3 | return !input_data.empty() && !input_data[0].empty(); |
| 40 | } | ||
| 41 | |||
| 42 | return true; | ||
| 43 | } | ||
| 44 | |||
| 45 | 6 | bool KapanovaSImageSmoothingMPI::PreProcessingImpl() { | |
| 46 | 6 | int rank = 0; | |
| 47 | 6 | MPI_Comm_rank(MPI_COMM_WORLD, &rank); | |
| 48 | |||
| 49 |
2/2✓ Branch 0 taken 3 times.
✓ Branch 1 taken 3 times.
|
6 | if (rank == 0) { |
| 50 | const auto &input_data = GetInput(); | ||
| 51 |
2/4✓ Branch 0 taken 3 times.
✗ Branch 1 not taken.
✓ Branch 2 taken 3 times.
✗ Branch 3 not taken.
|
3 | if (input_data.empty() || input_data[0].size() < 4) { |
| 52 | ✗ | return false; | |
| 53 | } | ||
| 54 | |||
| 55 | const auto &data = input_data[0]; | ||
| 56 | 3 | width_ = (data[1] << 8) | data[0]; | |
| 57 | 3 | height_ = (data[3] << 8) | data[2]; | |
| 58 | |||
| 59 | 3 | const auto width_u = static_cast<size_t>(width_); | |
| 60 | 3 | const auto height_u = static_cast<size_t>(height_); | |
| 61 | 3 | const size_t required_pixels = width_u * height_u * 3U; | |
| 62 | 3 | const size_t total_required_size = 4U + required_pixels; | |
| 63 | |||
| 64 |
1/2✓ Branch 0 taken 3 times.
✗ Branch 1 not taken.
|
3 | if (data.size() < total_required_size) { |
| 65 | return false; | ||
| 66 | } | ||
| 67 | |||
| 68 | 3 | input_.assign(data.begin() + 4, data.end()); | |
| 69 | |||
| 70 | 3 | std::array<int, 2> dimensions = {width_, height_}; | |
| 71 | 3 | MPI_Bcast(dimensions.data(), 2, MPI_INT, 0, MPI_COMM_WORLD); | |
| 72 | |||
| 73 | 3 | int image_size = static_cast<int>(input_.size()); | |
| 74 | 3 | MPI_Bcast(&image_size, 1, MPI_INT, 0, MPI_COMM_WORLD); | |
| 75 | |||
| 76 |
1/2✓ Branch 0 taken 3 times.
✗ Branch 1 not taken.
|
3 | if (image_size > 0) { |
| 77 | 3 | MPI_Bcast(input_.data(), image_size, MPI_UNSIGNED_CHAR, 0, MPI_COMM_WORLD); | |
| 78 | } | ||
| 79 | |||
| 80 | } else { | ||
| 81 | 3 | std::array<int, 2> dimensions = {0, 0}; | |
| 82 | 3 | MPI_Bcast(dimensions.data(), 2, MPI_INT, 0, MPI_COMM_WORLD); | |
| 83 | 3 | width_ = dimensions[0]; | |
| 84 | 3 | height_ = dimensions[1]; | |
| 85 | |||
| 86 | 3 | int image_size = 0; | |
| 87 | 3 | MPI_Bcast(&image_size, 1, MPI_INT, 0, MPI_COMM_WORLD); | |
| 88 | |||
| 89 |
1/2✓ Branch 0 taken 3 times.
✗ Branch 1 not taken.
|
3 | if (image_size > 0) { |
| 90 | 3 | input_.resize(static_cast<size_t>(image_size)); | |
| 91 | 3 | MPI_Bcast(input_.data(), image_size, MPI_UNSIGNED_CHAR, 0, MPI_COMM_WORLD); | |
| 92 | } | ||
| 93 | } | ||
| 94 | |||
| 95 | 6 | const auto width_u = static_cast<size_t>(width_); | |
| 96 | 6 | const auto height_u = static_cast<size_t>(height_); | |
| 97 | 6 | const size_t required_pixels = width_u * height_u * 3U; | |
| 98 | 6 | result_.resize(required_pixels); | |
| 99 | 6 | kernel_ = CreateKernel(); | |
| 100 | |||
| 101 | 6 | return true; | |
| 102 | } | ||
| 103 | |||
| 104 | 6 | std::vector<float> KapanovaSImageSmoothingMPI::CreateKernel() { | |
| 105 | const auto kernel_size = static_cast<size_t>(kKernelSize) * static_cast<size_t>(kKernelSize); | ||
| 106 | 6 | std::vector<float> kernel(kernel_size, 0.0F); | |
| 107 | float norm = 0.0F; | ||
| 108 | |||
| 109 |
2/2✓ Branch 0 taken 18 times.
✓ Branch 1 taken 6 times.
|
24 | for (int i = -kRadius; i <= kRadius; ++i) { |
| 110 |
2/2✓ Branch 0 taken 54 times.
✓ Branch 1 taken 18 times.
|
72 | for (int j = -kRadius; j <= kRadius; ++j) { |
| 111 | 54 | const int index = ((i + kRadius) * kKernelSize) + (j + kRadius); | |
| 112 | 54 | kernel[static_cast<size_t>(index)] = std::exp(-static_cast<float>((i * i) + (j * j)) / (2.0F * kSigmaSquared)); | |
| 113 | 54 | norm += kernel[static_cast<size_t>(index)]; | |
| 114 | } | ||
| 115 | } | ||
| 116 | |||
| 117 |
1/2✓ Branch 0 taken 6 times.
✗ Branch 1 not taken.
|
6 | if (norm > 0.0F) { |
| 118 |
2/2✓ Branch 0 taken 54 times.
✓ Branch 1 taken 6 times.
|
60 | for (auto &value : kernel) { |
| 119 | 54 | value /= norm; | |
| 120 | } | ||
| 121 | } | ||
| 122 | |||
| 123 | 6 | return kernel; | |
| 124 | } | ||
| 125 | |||
| 126 | 3 | void KapanovaSImageSmoothingMPI::ProcessBorderRows() { | |
| 127 |
2/2✓ Branch 0 taken 9 times.
✓ Branch 1 taken 3 times.
|
12 | for (int x_coord = 0; x_coord < width_; ++x_coord) { |
| 128 | 9 | const auto top_pos = static_cast<size_t>(x_coord) * 3U; | |
| 129 | 9 | SmoothPixel(&result_[top_pos], x_coord, 0); | |
| 130 | |||
| 131 | 9 | const auto bottom_pos = static_cast<size_t>((height_ - 1) * width_ * 3) + (static_cast<size_t>(x_coord) * 3U); | |
| 132 | 9 | SmoothPixel(&result_[bottom_pos], x_coord, height_ - 1); | |
| 133 | } | ||
| 134 | 3 | } | |
| 135 | |||
| 136 | ✗ | void KapanovaSImageSmoothingMPI::ProcessRowRange(int start_row, int num_rows) { | |
| 137 | ✗ | const int end_row = std::min(start_row + num_rows, height_ - 1); | |
| 138 | |||
| 139 | ✗ | for (int y_coord = start_row; y_coord < end_row; ++y_coord) { | |
| 140 | ✗ | for (int x_coord = 0; x_coord < width_; ++x_coord) { | |
| 141 | ✗ | const auto pos = static_cast<size_t>(y_coord * width_ * 3) + (static_cast<size_t>(x_coord) * 3U); | |
| 142 | ✗ | SmoothPixel(&result_[pos], x_coord, y_coord); | |
| 143 | } | ||
| 144 | } | ||
| 145 | ✗ | } | |
| 146 | |||
| 147 | 3 | void KapanovaSImageSmoothingMPI::SendImageData(int worker_rank, int row) { | |
| 148 | const int data_size = CalculateDataSize(row); | ||
| 149 | const int start_pos = CalculateStartPosition(row); | ||
| 150 | |||
| 151 |
1/2✓ Branch 0 taken 3 times.
✗ Branch 1 not taken.
|
3 | if (data_size > 0) { |
| 152 | 3 | MPI_Send(&input_[static_cast<size_t>(start_pos)], data_size, MPI_UNSIGNED_CHAR, worker_rank, kTagData, | |
| 153 | MPI_COMM_WORLD); | ||
| 154 | } | ||
| 155 | 3 | } | |
| 156 | |||
| 157 | ✗ | int KapanovaSImageSmoothingMPI::CalculateDataSize(int row) const { | |
| 158 |
2/8✗ Branch 0 not taken.
✗ Branch 1 not taken.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✓ Branch 4 taken 3 times.
✗ Branch 5 not taken.
✓ Branch 6 taken 3 times.
✗ Branch 7 not taken.
|
3 | if (row <= 0 || row >= height_ - 1) { |
| 159 | return 0; | ||
| 160 | } | ||
| 161 | |||
| 162 | 3 | return 3 * width_ * 3; | |
| 163 | } | ||
| 164 | |||
| 165 | ✗ | int KapanovaSImageSmoothingMPI::CalculateStartPosition(int row) const { | |
| 166 |
1/6✗ Branch 0 not taken.
✗ Branch 1 not taken.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✓ Branch 4 taken 3 times.
✗ Branch 5 not taken.
|
3 | if (row <= 0 || row >= height_ - 1) { |
| 167 | return 0; | ||
| 168 | } | ||
| 169 | |||
| 170 | 3 | return (row - 1) * width_ * 3; | |
| 171 | } | ||
| 172 | |||
| 173 | 3 | void KapanovaSImageSmoothingMPI::MasterProcess() { | |
| 174 | 3 | const int satellites = GetCommSize() - 1; | |
| 175 | |||
| 176 | 3 | ProcessBorderRows(); | |
| 177 | |||
| 178 | 3 | DistributeRowsToWorkers(satellites); | |
| 179 | |||
| 180 | 3 | SendExitSignalToWorkers(satellites); | |
| 181 | 3 | } | |
| 182 | |||
| 183 | 3 | void KapanovaSImageSmoothingMPI::DistributeRowsToWorkers(int num_workers) { | |
| 184 | int current_row = 1; | ||
| 185 | |||
| 186 |
2/2✓ Branch 0 taken 3 times.
✓ Branch 1 taken 3 times.
|
6 | while (current_row < height_ - 1) { |
| 187 | 3 | const int processes_to_use = std::min(num_workers, height_ - 1 - current_row); | |
| 188 | |||
| 189 | 3 | AssignRowsToWorkers(current_row, processes_to_use); | |
| 190 | 3 | ReceiveResultsFromWorkers(current_row, processes_to_use); | |
| 191 | 3 | current_row += processes_to_use; | |
| 192 | } | ||
| 193 | 3 | } | |
| 194 | |||
| 195 | 3 | void KapanovaSImageSmoothingMPI::AssignRowsToWorkers(int start_row, int num_workers) { | |
| 196 |
2/2✓ Branch 0 taken 3 times.
✓ Branch 1 taken 3 times.
|
6 | for (int i = 0; i < num_workers; ++i) { |
| 197 | 3 | const int worker_rank = i + 1; | |
| 198 | 3 | const int row_to_process = start_row + i; | |
| 199 | |||
| 200 |
2/4✓ Branch 0 taken 3 times.
✗ Branch 1 not taken.
✗ Branch 2 not taken.
✓ Branch 3 taken 3 times.
|
3 | if (row_to_process <= 0 || row_to_process >= height_ - 1) { |
| 201 | ✗ | continue; | |
| 202 | } | ||
| 203 | |||
| 204 | 3 | std::array<int, 2> work_info = {kNoEscapeSignal, row_to_process}; | |
| 205 | 3 | MPI_Send(work_info.data(), 2, MPI_INT, worker_rank, kTagExit, MPI_COMM_WORLD); | |
| 206 | |||
| 207 | 3 | SendImageData(worker_rank, row_to_process); | |
| 208 | } | ||
| 209 | 3 | } | |
| 210 | |||
| 211 | 3 | void KapanovaSImageSmoothingMPI::ReceiveResultsFromWorkers(int start_row, int num_workers) { | |
| 212 |
2/2✓ Branch 0 taken 3 times.
✓ Branch 1 taken 3 times.
|
6 | for (int i = 0; i < num_workers; ++i) { |
| 213 | 3 | const int worker_rank = i + 1; | |
| 214 | 3 | const int result_row = start_row + i; | |
| 215 | |||
| 216 |
2/4✓ Branch 0 taken 3 times.
✗ Branch 1 not taken.
✓ Branch 2 taken 3 times.
✗ Branch 3 not taken.
|
3 | if (result_row >= 1 && result_row < height_ - 1) { |
| 217 | 3 | const auto result_pos = static_cast<size_t>(result_row) * static_cast<size_t>(width_) * 3; | |
| 218 | 3 | MPI_Recv(&result_[result_pos], width_ * 3, MPI_UNSIGNED_CHAR, worker_rank, kTagResult, MPI_COMM_WORLD, | |
| 219 | MPI_STATUS_IGNORE); | ||
| 220 | } | ||
| 221 | } | ||
| 222 | 3 | } | |
| 223 | |||
| 224 | 3 | void KapanovaSImageSmoothingMPI::SendExitSignalToWorkers(int num_workers) { | |
| 225 |
2/2✓ Branch 0 taken 3 times.
✓ Branch 1 taken 3 times.
|
6 | for (int i = 1; i <= num_workers; ++i) { |
| 226 | 3 | std::array<int, 2> exit_info = {kEscapeSignal, 0}; | |
| 227 | 3 | MPI_Send(exit_info.data(), 2, MPI_INT, i, kTagExit, MPI_COMM_WORLD); | |
| 228 | } | ||
| 229 | 3 | } | |
| 230 | |||
| 231 | 3 | void KapanovaSImageSmoothingMPI::WorkerProcess() { | |
| 232 | 3 | int local_width = width_; | |
| 233 | |||
| 234 | 3 | std::vector<uint8_t> local_input; | |
| 235 |
1/4✓ Branch 1 taken 3 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
|
3 | std::vector<uint8_t> local_result(static_cast<size_t>(local_width * 3)); |
| 236 | |||
| 237 | 3 | std::array<int, 2> work_info = {0, 0}; | |
| 238 | |||
| 239 | while (true) { | ||
| 240 |
1/2✓ Branch 1 taken 6 times.
✗ Branch 2 not taken.
|
6 | MPI_Recv(work_info.data(), 2, MPI_INT, 0, kTagExit, MPI_COMM_WORLD, MPI_STATUS_IGNORE); |
| 241 | 6 | int escape_signal = work_info[0]; | |
| 242 | 6 | int row_to_process = work_info[1]; | |
| 243 | |||
| 244 |
2/2✓ Branch 0 taken 3 times.
✓ Branch 1 taken 3 times.
|
6 | if (escape_signal == kEscapeSignal) { |
| 245 | break; | ||
| 246 | } | ||
| 247 | |||
| 248 |
1/2✓ Branch 1 taken 3 times.
✗ Branch 2 not taken.
|
3 | const int received_bytes = ReceiveImageData(local_input); |
| 249 | 3 | const int rows_received = received_bytes / (local_width * 3); | |
| 250 | |||
| 251 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 3 times.
|
3 | if (rows_received > 0) { |
| 252 |
1/2✓ Branch 1 taken 3 times.
✗ Branch 2 not taken.
|
3 | ProcessAndSendResult(local_width, local_input, local_result, rows_received, row_to_process); |
| 253 | } | ||
| 254 | } | ||
| 255 | 3 | } | |
| 256 | |||
| 257 | 3 | int KapanovaSImageSmoothingMPI::ReceiveImageData(std::vector<uint8_t> &buffer) { | |
| 258 | MPI_Status status; | ||
| 259 | 3 | MPI_Probe(0, kTagData, MPI_COMM_WORLD, &status); | |
| 260 | |||
| 261 | 3 | int count = 0; | |
| 262 | 3 | MPI_Get_count(&status, MPI_UNSIGNED_CHAR, &count); | |
| 263 | |||
| 264 | 3 | buffer.resize(static_cast<size_t>(count)); | |
| 265 | 3 | MPI_Recv(buffer.data(), count, MPI_UNSIGNED_CHAR, 0, kTagData, MPI_COMM_WORLD, MPI_STATUS_IGNORE); | |
| 266 | |||
| 267 | 3 | return count; | |
| 268 | } | ||
| 269 | |||
| 270 | 3 | void KapanovaSImageSmoothingMPI::ProcessAndSendResult(int local_width, const std::vector<uint8_t> &input, | |
| 271 | std::vector<uint8_t> &result, int rows_received, | ||
| 272 | int row_to_process) { | ||
| 273 | const int target_row = row_to_process; | ||
| 274 | |||
| 275 |
2/2✓ Branch 0 taken 11 times.
✓ Branch 1 taken 3 times.
|
14 | for (int x_coord = 0; x_coord < local_width; ++x_coord) { |
| 276 | 11 | const auto pos = static_cast<size_t>(x_coord) * 3U; | |
| 277 | 11 | SmoothPixel(&result[pos], x_coord, target_row, true, &input, local_width, rows_received); | |
| 278 | } | ||
| 279 | |||
| 280 | 3 | MPI_Send(result.data(), local_width * 3, MPI_UNSIGNED_CHAR, 0, kTagResult, MPI_COMM_WORLD); | |
| 281 | 3 | } | |
| 282 | |||
| 283 | ✗ | int KapanovaSImageSmoothingMPI::GetCommRank() { | |
| 284 | 6 | int rank = 0; | |
| 285 | ✗ | MPI_Comm_rank(MPI_COMM_WORLD, &rank); | |
| 286 | 6 | return rank; | |
| 287 | } | ||
| 288 | |||
| 289 | ✗ | int KapanovaSImageSmoothingMPI::GetCommSize() { | |
| 290 | 9 | int size = 0; | |
| 291 | 9 | MPI_Comm_size(MPI_COMM_WORLD, &size); | |
| 292 | 3 | return size; | |
| 293 | } | ||
| 294 | |||
| 295 | 6 | bool KapanovaSImageSmoothingMPI::RunImpl() { | |
| 296 | int rank = GetCommRank(); | ||
| 297 | int size = GetCommSize(); | ||
| 298 | |||
| 299 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 6 times.
|
6 | if (size == 1) { |
| 300 | ✗ | ProcessRowRange(1, height_ - 2); | |
| 301 | ✗ | ProcessBorderRows(); | |
| 302 | ✗ | return true; | |
| 303 | } | ||
| 304 | |||
| 305 |
2/2✓ Branch 0 taken 3 times.
✓ Branch 1 taken 3 times.
|
6 | if (rank == 0) { |
| 306 | 3 | MasterProcess(); | |
| 307 | } else { | ||
| 308 | 3 | WorkerProcess(); | |
| 309 | } | ||
| 310 | |||
| 311 | 6 | MPI_Barrier(MPI_COMM_WORLD); | |
| 312 | |||
| 313 | 6 | return true; | |
| 314 | } | ||
| 315 | |||
| 316 | 6 | bool KapanovaSImageSmoothingMPI::PostProcessingImpl() { | |
| 317 | 6 | GetOutput() = result_; | |
| 318 | 6 | return true; | |
| 319 | } | ||
| 320 | |||
| 321 | // Вспомогательные функции для уменьшения когнитивной сложности SmoothPixel | ||
| 322 | namespace { | ||
| 323 | 99 | void ProcessLocalKernelPixel(int ry, int rx, int radius, int k_size, int local_width, int local_height, | |
| 324 | int local_y_center, int x_coord, const std::vector<uint8_t> &local_input, | ||
| 325 | const std::vector<float> &kernel, float &out_r, float &out_g, float &out_b) { | ||
| 326 | 99 | int local_y = local_y_center + ry; | |
| 327 | |||
| 328 |
1/2✓ Branch 0 taken 99 times.
✗ Branch 1 not taken.
|
99 | if (local_y < 0 || local_y >= local_height) { |
| 329 | return; | ||
| 330 | } | ||
| 331 | |||
| 332 | auto clamp = [](int n, int lo, int hi) { return std::min(std::max(n, lo), hi); }; | ||
| 333 | 99 | int local_x = clamp(x_coord + rx, 0, local_width - 1); | |
| 334 | |||
| 335 | const int local_stride = local_width * 3; | ||
| 336 | 99 | const auto pixel_pos = static_cast<size_t>(local_y * local_stride) + (static_cast<size_t>(local_x) * 3U); | |
| 337 | 99 | const auto kernel_pos = static_cast<size_t>((ry + radius) * k_size) + static_cast<size_t>(rx + radius); | |
| 338 | |||
| 339 | 99 | out_r += static_cast<float>(local_input[pixel_pos]) * kernel[kernel_pos]; | |
| 340 | 99 | out_g += static_cast<float>(local_input[pixel_pos + 1U]) * kernel[kernel_pos]; | |
| 341 | 99 | out_b += static_cast<float>(local_input[pixel_pos + 2U]) * kernel[kernel_pos]; | |
| 342 | } | ||
| 343 | |||
| 344 | 162 | void ProcessGlobalKernelPixel(int ry, int rx, int radius, int k_size, int width, int height, int x_coord, int y_coord, | |
| 345 | const std::vector<uint8_t> &input, const std::vector<float> &kernel, float &out_r, | ||
| 346 | float &out_g, float &out_b) { | ||
| 347 | auto clamp = [](int n, int lo, int hi) { return std::min(std::max(n, lo), hi); }; | ||
| 348 | 162 | int global_y = clamp(y_coord + ry, 0, height - 1); | |
| 349 | 162 | int global_x = clamp(x_coord + rx, 0, width - 1); | |
| 350 | |||
| 351 | 162 | const auto pixel_pos = static_cast<size_t>(global_y * width * 3) + (static_cast<size_t>(global_x) * 3U); | |
| 352 | 162 | const auto kernel_pos = static_cast<size_t>((ry + radius) * k_size) + static_cast<size_t>(rx + radius); | |
| 353 | |||
| 354 | 162 | out_r += static_cast<float>(input[pixel_pos]) * kernel[kernel_pos]; | |
| 355 | 162 | out_g += static_cast<float>(input[pixel_pos + 1U]) * kernel[kernel_pos]; | |
| 356 | 162 | out_b += static_cast<float>(input[pixel_pos + 2U]) * kernel[kernel_pos]; | |
| 357 | 162 | } | |
| 358 | |||
| 359 | 29 | void ApplyKernelToPixel(bool use_local, int radius, int width, int height, int x_coord, int y_coord, | |
| 360 | const std::vector<uint8_t> &global_input, const std::vector<float> &kernel, | ||
| 361 | const std::vector<uint8_t> *local_input, int local_width, int local_height, float &out_r, | ||
| 362 | float &out_g, float &out_b) { | ||
| 363 | 29 | const int k_size = (2 * radius) + 1; | |
| 364 | |||
| 365 |
2/2✓ Branch 0 taken 11 times.
✓ Branch 1 taken 18 times.
|
29 | if (use_local && local_input != nullptr) { |
| 366 | const int local_y_center = 1; | ||
| 367 | |||
| 368 |
2/2✓ Branch 0 taken 33 times.
✓ Branch 1 taken 11 times.
|
44 | for (int ry = -radius; ry <= radius; ++ry) { |
| 369 |
2/2✓ Branch 0 taken 99 times.
✓ Branch 1 taken 33 times.
|
132 | for (int rx = -radius; rx <= radius; ++rx) { |
| 370 | 99 | ProcessLocalKernelPixel(ry, rx, radius, k_size, local_width, local_height, local_y_center, x_coord, | |
| 371 | *local_input, kernel, out_r, out_g, out_b); | ||
| 372 | } | ||
| 373 | } | ||
| 374 | } else { | ||
| 375 |
2/2✓ Branch 0 taken 54 times.
✓ Branch 1 taken 18 times.
|
72 | for (int ry = -radius; ry <= radius; ++ry) { |
| 376 |
2/2✓ Branch 0 taken 162 times.
✓ Branch 1 taken 54 times.
|
216 | for (int rx = -radius; rx <= radius; ++rx) { |
| 377 | 162 | ProcessGlobalKernelPixel(ry, rx, radius, k_size, width, height, x_coord, y_coord, global_input, kernel, out_r, | |
| 378 | out_g, out_b); | ||
| 379 | } | ||
| 380 | } | ||
| 381 | } | ||
| 382 | 29 | } | |
| 383 | |||
| 384 | 29 | void SetPixelValue(uint8_t *out, float out_r, float out_g, float out_b) { | |
| 385 | 29 | out[0] = static_cast<uint8_t>(std::clamp(static_cast<int>(std::round(out_r)), 0, 255)); | |
| 386 | 29 | out[1] = static_cast<uint8_t>(std::clamp(static_cast<int>(std::round(out_g)), 0, 255)); | |
| 387 | 29 | out[2] = static_cast<uint8_t>(std::clamp(static_cast<int>(std::round(out_b)), 0, 255)); | |
| 388 | 29 | } | |
| 389 | } // namespace | ||
| 390 | |||
| 391 | 29 | void KapanovaSImageSmoothingMPI::SmoothPixel(uint8_t *out, int x_coord, int y_coord, bool use_local, | |
| 392 | const std::vector<uint8_t> *local_input, int local_width, | ||
| 393 | int local_height) { | ||
| 394 | 29 | float out_r = 0.0F; | |
| 395 | 29 | float out_g = 0.0F; | |
| 396 | 29 | float out_b = 0.0F; | |
| 397 | |||
| 398 | 29 | ApplyKernelToPixel(use_local, radius_, width_, height_, x_coord, y_coord, input_, kernel_, local_input, local_width, | |
| 399 | local_height, out_r, out_g, out_b); | ||
| 400 | |||
| 401 | 29 | SetPixelValue(out, out_r, out_g, out_b); | |
| 402 | 29 | } | |
| 403 | |||
| 404 | } // namespace kapanova_s_image_smoothing | ||
| 405 |