| Line | Branch | Exec | Source |
|---|---|---|---|
| 1 | #include "../include/rect_method_all.hpp" | ||
| 2 | |||
| 3 | #include <mpi.h> | ||
| 4 | #include <omp.h> | ||
| 5 | |||
| 6 | #include <algorithm> | ||
| 7 | #include <cmath> | ||
| 8 | #include <cstddef> | ||
| 9 | #include <functional> | ||
| 10 | #include <utility> | ||
| 11 | #include <vector> | ||
| 12 | |||
| 13 | #include "../../common/include/common.hpp" | ||
| 14 | #include "util/include/util.hpp" | ||
| 15 | |||
| 16 | namespace kutergin_v_multidimensional_integration_rect_method { | ||
| 17 | |||
| 18 |
1/2✓ Branch 1 taken 10 times.
✗ Branch 2 not taken.
|
10 | RectMethodALL::RectMethodALL(const InType &in) { |
| 19 | SetTypeOfTask(GetStaticTypeOfTask()); | ||
| 20 |
1/2✓ Branch 1 taken 10 times.
✗ Branch 2 not taken.
|
10 | GetInput() = in; |
| 21 | 10 | GetOutput() = 0.0; | |
| 22 | 10 | } | |
| 23 | |||
| 24 |
1/2✓ Branch 0 taken 10 times.
✗ Branch 1 not taken.
|
10 | bool RectMethodALL::ValidationImpl() { |
| 25 | const auto &input = GetInput(); | ||
| 26 |
2/4✓ Branch 0 taken 10 times.
✗ Branch 1 not taken.
✓ Branch 2 taken 10 times.
✗ Branch 3 not taken.
|
10 | if (input.limits.size() != input.n_steps.size() || input.limits.empty()) { |
| 27 | return false; | ||
| 28 | } | ||
| 29 | return std::ranges::all_of(input.n_steps, [](int n) { return n > 0; }); | ||
| 30 | } | ||
| 31 | |||
| 32 | 10 | bool RectMethodALL::PreProcessingImpl() { | |
| 33 | 10 | local_input_ = GetInput(); | |
| 34 | 10 | res_ = 0.0; | |
| 35 | 10 | return true; | |
| 36 | } | ||
| 37 | |||
| 38 | 10 | bool RectMethodALL::RunImpl() { | |
| 39 | 10 | int rank = 0; | |
| 40 | 10 | int size = 1; | |
| 41 | 10 | bool is_mpi = ppc::util::IsUnderMpirun(); | |
| 42 | |||
| 43 |
1/2✓ Branch 0 taken 10 times.
✗ Branch 1 not taken.
|
10 | if (is_mpi) { |
| 44 | 10 | MPI_Comm_rank(MPI_COMM_WORLD, &rank); | |
| 45 | 10 | MPI_Comm_size(MPI_COMM_WORLD, &size); | |
| 46 | } | ||
| 47 | |||
| 48 |
2/2✓ Branch 0 taken 5 times.
✓ Branch 1 taken 5 times.
|
10 | size_t dims = (rank == 0) ? local_input_.limits.size() : 0; |
| 49 | |||
| 50 | 10 | DistributeData(rank, dims); | |
| 51 | |||
| 52 | size_t total_iterations = 1; | ||
| 53 | 10 | std::vector<double> h(dims); | |
| 54 | double d_v = 1.0; | ||
| 55 | |||
| 56 |
2/2✓ Branch 0 taken 30 times.
✓ Branch 1 taken 10 times.
|
40 | for (size_t i = 0; i < dims; ++i) { |
| 57 | 30 | total_iterations *= local_input_.n_steps[i]; | |
| 58 | 30 | h[i] = (local_input_.limits[i].second - local_input_.limits[i].first) / local_input_.n_steps[i]; | |
| 59 | 30 | d_v *= h[i]; | |
| 60 | } | ||
| 61 | |||
| 62 | 10 | size_t proc_chunk = total_iterations / size; | |
| 63 | 10 | size_t proc_remainder = total_iterations % size; | |
| 64 | |||
| 65 |
1/2✓ Branch 0 taken 10 times.
✗ Branch 1 not taken.
|
10 | size_t my_proc_count = proc_chunk + (std::cmp_less(rank, proc_remainder) ? 1 : 0); |
| 66 |
1/2✓ Branch 0 taken 10 times.
✗ Branch 1 not taken.
|
10 | size_t my_proc_start = (rank * proc_chunk) + std::min(static_cast<size_t>(rank), proc_remainder); |
| 67 | |||
| 68 | 10 | double local_sum = 0.0; | |
| 69 | |||
| 70 |
1/2✓ Branch 0 taken 10 times.
✗ Branch 1 not taken.
|
10 | if (my_proc_count > 0) { |
| 71 |
1/2✓ Branch 1 taken 10 times.
✗ Branch 2 not taken.
|
10 | int num_threads = ppc::util::GetNumThreads(); |
| 72 | 10 | omp_set_num_threads(num_threads); | |
| 73 | |||
| 74 | 10 | const auto &func = local_input_.func; | |
| 75 | 10 | const auto &n_steps = local_input_.n_steps; | |
| 76 | 10 | const auto &limits = local_input_.limits; | |
| 77 | |||
| 78 | 10 | #pragma omp parallel default(none) shared(h, dims, my_proc_start, my_proc_count, func, n_steps, limits) \ | |
| 79 | reduction(+ : local_sum) | ||
| 80 | { | ||
| 81 | int tid = omp_get_thread_num(); | ||
| 82 | int t_count = omp_get_num_threads(); | ||
| 83 | |||
| 84 | size_t thread_chunk = my_proc_count / t_count; | ||
| 85 | size_t thread_remainder = my_proc_count % t_count; | ||
| 86 | |||
| 87 | size_t my_thread_count = thread_chunk + (std::cmp_less(tid, thread_remainder) ? 1 : 0); | ||
| 88 | size_t my_thread_start = | ||
| 89 | my_proc_start + (tid * thread_chunk) + std::min(static_cast<size_t>(tid), thread_remainder); | ||
| 90 | |||
| 91 | local_sum += CalculateChunkSum(my_thread_start, my_thread_start + my_thread_count, h, limits, n_steps, func); | ||
| 92 | } | ||
| 93 | } | ||
| 94 | |||
| 95 |
1/2✓ Branch 0 taken 10 times.
✗ Branch 1 not taken.
|
10 | if (is_mpi) { |
| 96 | 10 | double global_sum = 0.0; | |
| 97 |
1/2✓ Branch 1 taken 10 times.
✗ Branch 2 not taken.
|
10 | MPI_Reduce(&local_sum, &global_sum, 1, MPI_DOUBLE, MPI_SUM, 0, MPI_COMM_WORLD); |
| 98 |
2/2✓ Branch 0 taken 5 times.
✓ Branch 1 taken 5 times.
|
10 | if (rank == 0) { |
| 99 | 5 | res_ = global_sum * d_v; | |
| 100 | } | ||
| 101 | } else { | ||
| 102 | ✗ | res_ = local_sum * d_v; | |
| 103 | } | ||
| 104 | |||
| 105 | 10 | return true; | |
| 106 | } | ||
| 107 | |||
| 108 | 10 | bool RectMethodALL::PostProcessingImpl() { | |
| 109 | 10 | int rank = 0; | |
| 110 |
1/2✓ Branch 1 taken 10 times.
✗ Branch 2 not taken.
|
10 | if (ppc::util::IsUnderMpirun()) { |
| 111 | 10 | MPI_Comm_rank(MPI_COMM_WORLD, &rank); | |
| 112 | } | ||
| 113 |
2/2✓ Branch 0 taken 5 times.
✓ Branch 1 taken 5 times.
|
10 | if (rank == 0) { |
| 114 | 5 | GetOutput() = res_; | |
| 115 | } | ||
| 116 | 10 | return true; | |
| 117 | } | ||
| 118 | |||
| 119 | 10 | void RectMethodALL::DistributeData(int rank, size_t &dims) { | |
| 120 |
1/2✗ Branch 1 not taken.
✓ Branch 2 taken 10 times.
|
10 | if (!ppc::util::IsUnderMpirun()) { |
| 121 | ✗ | return; | |
| 122 | } | ||
| 123 | |||
| 124 | 10 | int dims_io = static_cast<int>(dims); | |
| 125 | 10 | MPI_Bcast(&dims_io, 1, MPI_INT, 0, MPI_COMM_WORLD); | |
| 126 | 10 | dims = static_cast<size_t>(dims_io); | |
| 127 | |||
| 128 |
2/2✓ Branch 0 taken 5 times.
✓ Branch 1 taken 5 times.
|
10 | if (rank != 0) { |
| 129 | 5 | local_input_.limits.resize(dims); | |
| 130 | 5 | local_input_.n_steps.resize(dims); | |
| 131 | } | ||
| 132 |
2/2✓ Branch 0 taken 30 times.
✓ Branch 1 taken 10 times.
|
40 | for (size_t i = 0; i < dims; ++i) { |
| 133 | 30 | MPI_Bcast(&local_input_.limits[i].first, 1, MPI_DOUBLE, 0, MPI_COMM_WORLD); | |
| 134 | 30 | MPI_Bcast(&local_input_.limits[i].second, 1, MPI_DOUBLE, 0, MPI_COMM_WORLD); | |
| 135 | 30 | MPI_Bcast(&local_input_.n_steps[i], 1, MPI_INT, 0, MPI_COMM_WORLD); | |
| 136 | } | ||
| 137 | } | ||
| 138 | |||
| 139 | 20 | double RectMethodALL::CalculateChunkSum(size_t start_idx, size_t end_idx, const std::vector<double> &h, | |
| 140 | const std::vector<std::pair<double, double>> &limits, | ||
| 141 | const std::vector<int> &n_steps, | ||
| 142 | const std::function<double(const std::vector<double> &)> &func) { | ||
| 143 |
1/2✓ Branch 0 taken 20 times.
✗ Branch 1 not taken.
|
20 | if (start_idx >= end_idx) { |
| 144 | return 0.0; | ||
| 145 | } | ||
| 146 | |||
| 147 | 20 | size_t count = end_idx - start_idx; | |
| 148 | size_t dims = limits.size(); // число размерностей пространства | ||
| 149 | 20 | std::vector<int> current_indices(dims, 0); | |
| 150 |
1/4✓ Branch 1 taken 20 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
|
20 | std::vector<double> coords(dims); // создание вектора координат размером dims |
| 151 | |||
| 152 | size_t temp_idx = start_idx; | ||
| 153 |
2/2✓ Branch 0 taken 60 times.
✓ Branch 1 taken 20 times.
|
80 | for (int dim = static_cast<int>(dims) - 1; dim >= 0; --dim) { |
| 154 | 60 | current_indices[dim] = static_cast<int>(temp_idx % n_steps[dim]); | |
| 155 | 60 | temp_idx /= n_steps[dim]; | |
| 156 | } | ||
| 157 | |||
| 158 | double chunk_sum = 0.0; | ||
| 159 | |||
| 160 |
2/2✓ Branch 0 taken 11468 times.
✓ Branch 1 taken 20 times.
|
11488 | for (size_t i = 0; i < count; ++i) { |
| 161 |
2/2✓ Branch 0 taken 32815 times.
✓ Branch 1 taken 11468 times.
|
44283 | for (size_t dm = 0; dm < dims; ++dm) { |
| 162 | 32815 | coords[dm] = limits[dm].first + ((current_indices[dm] + 0.5) * h[dm]); // реальная координата | |
| 163 | } | ||
| 164 | |||
| 165 | 11468 | chunk_sum += func(coords); // вычисление функции в точке | |
| 166 | |||
| 167 |
2/2✓ Branch 0 taken 12213 times.
✓ Branch 1 taken 5 times.
|
12218 | for (int dm = static_cast<int>(dims) - 1; dm >= 0; --dm) { |
| 168 |
2/2✓ Branch 0 taken 750 times.
✓ Branch 1 taken 11463 times.
|
12213 | current_indices[dm]++; |
| 169 |
2/2✓ Branch 0 taken 750 times.
✓ Branch 1 taken 11463 times.
|
12213 | if (current_indices[dm] < n_steps[dm]) { |
| 170 | break; | ||
| 171 | } | ||
| 172 | 750 | current_indices[dm] = 0; | |
| 173 | } | ||
| 174 | } | ||
| 175 | |||
| 176 | return chunk_sum; | ||
| 177 | } | ||
| 178 | |||
| 179 | } // namespace kutergin_v_multidimensional_integration_rect_method | ||
| 180 |