Болтенков Станислав. Технология OMP. Линейная фильтрация изображений (вертикальное разбиение). Ядро Гаусса 3x3. Вариант 25.

tasks/boltenkov_s_gaussian_kernel/omp/include/ops_omp.hpp

@@ -0,0 +1,26 @@
+#pragma once
+
+#include <vector>
+
+#include "boltenkov_s_gaussian_kernel/common/include/common.hpp"
+#include "task/include/task.hpp"
+
+namespace boltenkov_s_gaussian_kernel {
+
+class BoltenkovSGaussianKernelOMP : public BaseTask {
+ public:
+  static constexpr ppc::task::TypeOfTask GetStaticTypeOfTask() {
+    return ppc::task::TypeOfTask::kOMP;
+  }
+  explicit BoltenkovSGaussianKernelOMP(const InType &in);
+
+ private:
+  std::vector<std::vector<int>> kernel_;
+  int shift_ = 4;
+  bool ValidationImpl() override;
+  bool PreProcessingImpl() override;
+  bool RunImpl() override;
+  bool PostProcessingImpl() override;
+};
+
+}  // namespace boltenkov_s_gaussian_kernel

tasks/boltenkov_s_gaussian_kernel/omp/src/ops_omp.cpp

@@ -0,0 +1,79 @@
+#include "boltenkov_s_gaussian_kernel/omp/include/ops_omp.hpp"
+
+#include <omp.h>
+
+#include <algorithm>
+#include <cstddef>
+#include <vector>
+
+#include "boltenkov_s_gaussian_kernel/common/include/common.hpp"
+#include "util/include/util.hpp"
+
+namespace boltenkov_s_gaussian_kernel {
+
+BoltenkovSGaussianKernelOMP::BoltenkovSGaussianKernelOMP(const InType &in)
+    : kernel_{{{1, 2, 1}, {2, 4, 2}, {1, 2, 1}}} {
+  SetTypeOfTask(GetStaticTypeOfTask());
+  GetInput() = in;
+  GetOutput() = std::vector<std::vector<int>>();
+}
+
+bool BoltenkovSGaussianKernelOMP::ValidationImpl() {
+  std::size_t n = std::get<0>(GetInput());
+  std::size_t m = std::get<1>(GetInput());
+  if (std::get<2>(GetInput()).size() != n) {
+    return false;
+  }
+  for (std::size_t i = 0; i < n; i++) {
+    if (std::get<2>(GetInput())[i].size() != m) {
+      return false;
+    }
+  }
+  return true;
+}
+
+bool BoltenkovSGaussianKernelOMP::PreProcessingImpl() {
+  GetOutput().resize(std::get<0>(GetInput()));
+  for (std::size_t i = 0; i < std::get<0>(GetInput()); i++) {
+    GetOutput()[i].resize(std::get<1>(GetInput()));
+  }
+  return true;
+}
+
+bool BoltenkovSGaussianKernelOMP::RunImpl() {
+  std::size_t n = std::get<0>(GetInput());
+  std::size_t m = std::get<1>(GetInput());
+
+  std::vector<std::vector<int>> data = std::get<2>(GetInput());
+  std::vector<std::vector<int>> tmp_data(n + 2, std::vector<int>(m + 2, 0));
+  std::vector<std::vector<int>> &res = GetOutput();
+
+#pragma omp parallel for num_threads(ppc::util::GetNumThreads()) default(none) shared(tmp_data, data) firstprivate(n)
+  for (std::size_t i = 1; i <= n; i++) {
+    std::copy(data[i - 1].begin(), data[i - 1].end(), tmp_data[i].begin() + 1);
+  }
+
+  auto kernel = kernel_;
+  int shift = shift_;
+
+#pragma omp parallel for num_threads(ppc::util::GetNumThreads()) default(none) shared(tmp_data, res) \
+    firstprivate(n, m, kernel, shift)
+  for (std::size_t i = 1; i <= n; i++) {
+    for (std::size_t j = 1; j <= m; j++) {
+      res[i - 1][j - 1] = (tmp_data[i - 1][j - 1] * kernel[0][0]) + (tmp_data[i - 1][j] * kernel[0][1]) +
+                          (tmp_data[i - 1][j + 1] * kernel[0][2]) + (tmp_data[i][j - 1] * kernel[1][0]) +
+                          (tmp_data[i][j] * kernel[1][1]) + (tmp_data[i][j + 1] * kernel[1][2]) +
+                          (tmp_data[i + 1][j - 1] * kernel[2][0]) + (tmp_data[i + 1][j] * kernel[2][1]) +
+                          (tmp_data[i + 1][j + 1] * kernel[2][2]);
+      res[i - 1][j - 1] >>= shift;
+    }
+  }
+
+  return true;
+}
+
+bool BoltenkovSGaussianKernelOMP::PostProcessingImpl() {
+  return true;
+}
+
+}  // namespace boltenkov_s_gaussian_kernel

tasks/boltenkov_s_gaussian_kernel/tests/functional/main.cpp

@@ -12,6 +12,7 @@
 #include <vector>
 
 #include "boltenkov_s_gaussian_kernel/common/include/common.hpp"
+#include "boltenkov_s_gaussian_kernel/omp/include/ops_omp.hpp"
 #include "boltenkov_s_gaussian_kernel/seq/include/ops_seq.hpp"
 #include "util/include/func_test_util.hpp"
 #include "util/include/util.hpp"
@@ -100,7 +101,8 @@ TEST_P(BoltenkovSRunFuncTestsProcesses, MatmulFromPic) {
 const std::array<TestType, 2> kTestParam = {"pic1", "pic2"};
 
 const auto kTestTasksList = std::tuple_cat(
-    ppc::util::AddFuncTask<BoltenkovSGaussianKernelSEQ, InType>(kTestParam, PPC_SETTINGS_boltenkov_s_gaussian_kernel));
+    ppc::util::AddFuncTask<BoltenkovSGaussianKernelSEQ, InType>(kTestParam, PPC_SETTINGS_boltenkov_s_gaussian_kernel),
+    ppc::util::AddFuncTask<BoltenkovSGaussianKernelOMP, InType>(kTestParam, PPC_SETTINGS_boltenkov_s_gaussian_kernel));
 
 const auto kGtestValues = ppc::util::ExpandToValues(kTestTasksList);
 

tasks/boltenkov_s_gaussian_kernel/tests/performance/main.cpp

@@ -9,6 +9,7 @@
 #include <vector>
 
 #include "boltenkov_s_gaussian_kernel/common/include/common.hpp"
+#include "boltenkov_s_gaussian_kernel/omp/include/ops_omp.hpp"
 #include "boltenkov_s_gaussian_kernel/seq/include/ops_seq.hpp"
 #include "util/include/perf_test_util.hpp"
 #include "util/include/util.hpp"
@@ -76,7 +77,8 @@ TEST_P(BoltenkovSRunPerfTestProcesses, RunPerfModes) {
 namespace {
 
 const auto kAllPerfTasks =
-    ppc::util::MakeAllPerfTasks<InType, BoltenkovSGaussianKernelSEQ>(PPC_SETTINGS_boltenkov_s_gaussian_kernel);
+    ppc::util::MakeAllPerfTasks<InType, BoltenkovSGaussianKernelSEQ, BoltenkovSGaussianKernelOMP>(
+        PPC_SETTINGS_boltenkov_s_gaussian_kernel);
 
 const auto kGtestValues = ppc::util::TupleToGTestValues(kAllPerfTasks);