| Line | Branch | Exec | Source |
|---|---|---|---|
| 1 | #include "chernykh_s_yadro_gaussa_horizontal/mpi/include/ops_mpi.hpp" | ||
| 2 | |||
| 3 | #include <mpi.h> | ||
| 4 | |||
| 5 | #include <algorithm> | ||
| 6 | #include <cstddef> | ||
| 7 | #include <vector> | ||
| 8 | |||
| 9 | #include "chernykh_s_yadro_gaussa_horizontal/common/include/common.hpp" | ||
| 10 | namespace chernykh_s_yadro_gaussa_horizontal { | ||
| 11 | |||
| 12 | 1238600 | int ChernykhSYadroGaussaHorizontalMPI::CalculateGauss(const std::vector<int> &data, int r, int c, int stolbci, | |
| 13 | int extended_rows) { | ||
| 14 | int pixel_sum = 0; | ||
| 15 |
2/2✓ Branch 0 taken 3715800 times.
✓ Branch 1 taken 1238600 times.
|
4954400 | for (int ki = -1; ki <= 1; ++ki) { |
| 16 |
2/2✓ Branch 0 taken 11147400 times.
✓ Branch 1 taken 3715800 times.
|
14863200 | for (int kj = -1; kj <= 1; ++kj) { |
| 17 | int weight = 0; | ||
| 18 |
2/2✓ Branch 0 taken 9908800 times.
✓ Branch 1 taken 1238600 times.
|
11147400 | if (ki == 0 && kj == 0) { |
| 19 | weight = 4; | ||
| 20 |
2/2✓ Branch 0 taken 4954400 times.
✓ Branch 1 taken 4954400 times.
|
9908800 | } else if (ki == 0 || kj == 0) { |
| 21 | weight = 2; | ||
| 22 | } else { | ||
| 23 | weight = 1; | ||
| 24 | } | ||
| 25 | |||
| 26 | 11147400 | int cur_r = std::clamp(r + ki, 0, extended_rows - 1); | |
| 27 | 11147400 | int cur_c = std::clamp(c + kj, 0, stolbci - 1); | |
| 28 | 11147400 | size_t idx = (static_cast<size_t>(cur_r) * static_cast<size_t>(stolbci)) + static_cast<size_t>(cur_c); | |
| 29 | 11147400 | pixel_sum += data[idx] * weight; | |
| 30 | } | ||
| 31 | } | ||
| 32 | 1238600 | return pixel_sum / 16; | |
| 33 | } | ||
| 34 | |||
| 35 | 10 | void ChernykhSYadroGaussaHorizontalMPI::GaussFilter(const std::vector<int> &local_data, std::vector<int> &local_res, | |
| 36 | int stroki_local, int stolbci, int halo_top) { | ||
| 37 | 10 | int extended_rows = static_cast<int>(local_data.size() / static_cast<size_t>(stolbci)); | |
| 38 |
2/2✓ Branch 0 taken 1914 times.
✓ Branch 1 taken 10 times.
|
1924 | for (int i = 0; i < stroki_local; ++i) { |
| 39 |
2/2✓ Branch 0 taken 1238600 times.
✓ Branch 1 taken 1914 times.
|
1240514 | for (int j = 0; j < stolbci; ++j) { |
| 40 | 1238600 | size_t res_idx = (static_cast<size_t>(i) * static_cast<size_t>(stolbci)) + static_cast<size_t>(j); | |
| 41 | 1238600 | local_res[res_idx] = CalculateGauss(local_data, i + halo_top, j, stolbci, extended_rows); | |
| 42 | } | ||
| 43 | } | ||
| 44 | 10 | } | |
| 45 | |||
| 46 |
1/2✓ Branch 1 taken 10 times.
✗ Branch 2 not taken.
|
10 | ChernykhSYadroGaussaHorizontalMPI::ChernykhSYadroGaussaHorizontalMPI(const InType &in) { |
| 47 | SetTypeOfTask(GetStaticTypeOfTask()); | ||
| 48 | 10 | GetInput() = InType(in); | |
| 49 | 10 | GetOutput() = std::vector<int>(); | |
| 50 | 10 | } | |
| 51 | |||
| 52 | 10 | bool ChernykhSYadroGaussaHorizontalMPI::ValidationImpl() { | |
| 53 | 10 | int stolbci = std::get<0>(GetInput()); | |
| 54 | 10 | int stroki = std::get<1>(GetInput()); | |
| 55 | auto &input = std::get<2>(GetInput()); | ||
| 56 |
2/4✓ Branch 0 taken 10 times.
✗ Branch 1 not taken.
✗ Branch 2 not taken.
✓ Branch 3 taken 10 times.
|
10 | return stolbci > 0 && stroki > 0 && input.size() == static_cast<size_t>(stroki) * static_cast<size_t>(stolbci); |
| 57 | } | ||
| 58 | |||
| 59 | 10 | bool ChernykhSYadroGaussaHorizontalMPI::PreProcessingImpl() { | |
| 60 | 10 | return true; | |
| 61 | } | ||
| 62 | |||
| 63 | 10 | bool ChernykhSYadroGaussaHorizontalMPI::RunImpl() { | |
| 64 | 10 | int rank = 0; | |
| 65 | 10 | int size = 0; | |
| 66 | 10 | MPI_Comm_rank(MPI_COMM_WORLD, &rank); | |
| 67 | 10 | MPI_Comm_size(MPI_COMM_WORLD, &size); | |
| 68 | |||
| 69 | 10 | int stolbci = 0; | |
| 70 | 10 | int stroki = 0; | |
| 71 | |||
| 72 |
2/2✓ Branch 0 taken 5 times.
✓ Branch 1 taken 5 times.
|
10 | if (rank == 0) { |
| 73 | 5 | stolbci = std::get<0>(GetInput()); | |
| 74 | 5 | stroki = std::get<1>(GetInput()); | |
| 75 | } | ||
| 76 | |||
| 77 | 10 | MPI_Bcast(&stolbci, 1, MPI_INT, 0, MPI_COMM_WORLD); | |
| 78 | 10 | MPI_Bcast(&stroki, 1, MPI_INT, 0, MPI_COMM_WORLD); | |
| 79 | |||
| 80 | 10 | std::vector<int> strok_na_process(size, 0); | |
| 81 |
1/4✓ Branch 1 taken 10 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
|
10 | std::vector<int> offset(size, 0); |
| 82 | int current_offset = 0; | ||
| 83 | |||
| 84 |
2/2✓ Branch 0 taken 20 times.
✓ Branch 1 taken 10 times.
|
30 | for (int i = 0; i < size; ++i) { |
| 85 |
1/2✓ Branch 0 taken 20 times.
✗ Branch 1 not taken.
|
20 | int count = (stroki / size) + (i < (stroki % size) ? 1 : 0); |
| 86 | 20 | strok_na_process[i] = count; | |
| 87 | 20 | offset[i] = current_offset; | |
| 88 | 20 | current_offset += count; | |
| 89 | } | ||
| 90 | |||
| 91 |
1/4✓ Branch 1 taken 10 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
|
10 | std::vector<int> sendcounts(size, 0); |
| 92 |
1/4✓ Branch 1 taken 10 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
|
10 | std::vector<int> new_start(size, 0); |
| 93 | |||
| 94 |
2/2✓ Branch 0 taken 20 times.
✓ Branch 1 taken 10 times.
|
30 | for (int i = 0; i < size; ++i) { |
| 95 |
2/2✓ Branch 0 taken 10 times.
✓ Branch 1 taken 10 times.
|
20 | int start_row = offset[i]; |
| 96 | 20 | int count_rows = strok_na_process[i]; | |
| 97 |
2/2✓ Branch 0 taken 10 times.
✓ Branch 1 taken 10 times.
|
20 | int h_top = (start_row > 0) ? 1 : 0; |
| 98 |
2/2✓ Branch 0 taken 10 times.
✓ Branch 1 taken 10 times.
|
20 | int h_bot = ((start_row + count_rows) < stroki) ? 1 : 0; |
| 99 | 20 | sendcounts[i] = (count_rows + h_top + h_bot) * stolbci; | |
| 100 | 20 | new_start[i] = (start_row - h_top) * stolbci; | |
| 101 | } | ||
| 102 | |||
| 103 |
2/2✓ Branch 0 taken 5 times.
✓ Branch 1 taken 5 times.
|
10 | int local_rows = strok_na_process[rank]; |
| 104 |
2/2✓ Branch 0 taken 5 times.
✓ Branch 1 taken 5 times.
|
10 | int halo_top = (offset[rank] > 0) ? 1 : 0; |
| 105 |
2/2✓ Branch 0 taken 5 times.
✓ Branch 1 taken 5 times.
|
10 | int halo_bot = ((offset[rank] + local_rows) < stroki) ? 1 : 0; |
| 106 | 10 | int extended_rows = local_rows + halo_top + halo_bot; | |
| 107 | |||
| 108 |
1/4✓ Branch 1 taken 10 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
|
10 | std::vector<int> local_data(static_cast<size_t>(extended_rows) * static_cast<size_t>(stolbci)); |
| 109 | |||
| 110 | const int *data_to_send = nullptr; | ||
| 111 | int *send_counts_ptr = nullptr; | ||
| 112 | int *new_start_ptr = nullptr; | ||
| 113 | |||
| 114 |
2/2✓ Branch 0 taken 5 times.
✓ Branch 1 taken 5 times.
|
10 | if (rank == 0) { |
| 115 | data_to_send = std::get<2>(GetInput()).data(); | ||
| 116 | send_counts_ptr = sendcounts.data(); | ||
| 117 | new_start_ptr = new_start.data(); | ||
| 118 | } | ||
| 119 | |||
| 120 |
1/2✓ Branch 1 taken 10 times.
✗ Branch 2 not taken.
|
10 | MPI_Scatterv(data_to_send, send_counts_ptr, new_start_ptr, MPI_INT, local_data.data(), extended_rows * stolbci, |
| 121 | MPI_INT, 0, MPI_COMM_WORLD); | ||
| 122 | |||
| 123 |
1/4✓ Branch 1 taken 10 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
|
10 | std::vector<int> local_res(static_cast<size_t>(local_rows) * static_cast<size_t>(stolbci)); |
| 124 | 10 | GaussFilter(local_data, local_res, local_rows, stolbci, halo_top); | |
| 125 | |||
| 126 | int *final_output_ptr = nullptr; | ||
| 127 |
1/4✓ Branch 1 taken 10 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
|
10 | std::vector<int> recvcounts(size, 0); |
| 128 |
1/4✓ Branch 1 taken 10 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
|
10 | std::vector<int> recvdispls(size, 0); |
| 129 | int *recvcounts_ptr = nullptr; | ||
| 130 | int *recvdispls_ptr = nullptr; | ||
| 131 | |||
| 132 |
2/2✓ Branch 0 taken 5 times.
✓ Branch 1 taken 5 times.
|
10 | if (rank == 0) { |
| 133 |
1/2✓ Branch 1 taken 5 times.
✗ Branch 2 not taken.
|
5 | GetOutput().resize(static_cast<size_t>(stolbci) * static_cast<size_t>(stroki)); |
| 134 | final_output_ptr = GetOutput().data(); | ||
| 135 |
2/2✓ Branch 0 taken 10 times.
✓ Branch 1 taken 5 times.
|
15 | for (int i = 0; i < size; ++i) { |
| 136 | 10 | recvcounts[i] = strok_na_process[i] * stolbci; | |
| 137 | 10 | recvdispls[i] = offset[i] * stolbci; | |
| 138 | } | ||
| 139 | recvcounts_ptr = recvcounts.data(); | ||
| 140 | recvdispls_ptr = recvdispls.data(); | ||
| 141 | } | ||
| 142 | |||
| 143 |
1/2✓ Branch 1 taken 10 times.
✗ Branch 2 not taken.
|
10 | MPI_Gatherv(local_res.data(), local_rows * stolbci, MPI_INT, final_output_ptr, recvcounts_ptr, recvdispls_ptr, |
| 144 | MPI_INT, 0, MPI_COMM_WORLD); | ||
| 145 | |||
| 146 | 10 | return true; | |
| 147 | } | ||
| 148 | |||
| 149 | 10 | bool ChernykhSYadroGaussaHorizontalMPI::PostProcessingImpl() { | |
| 150 | 10 | return true; | |
| 151 | } | ||
| 152 | } // namespace chernykh_s_yadro_gaussa_horizontal | ||
| 153 |