| Line | Branch | Exec | Source |
|---|---|---|---|
| 1 | #include "frolova_s_sum_elem_matrix/mpi/include/ops_mpi.hpp" | ||
| 2 | |||
| 3 | #include <mpi.h> | ||
| 4 | |||
| 5 | #include <limits> | ||
| 6 | #include <tuple> | ||
| 7 | #include <vector> | ||
| 8 | |||
| 9 | #include "frolova_s_sum_elem_matrix/common/include/common.hpp" | ||
| 10 | |||
| 11 | namespace frolova_s_sum_elem_matrix { | ||
| 12 | |||
| 13 |
1/2✓ Branch 1 taken 20 times.
✗ Branch 2 not taken.
|
20 | FrolovaSSumElemMatrixMPI::FrolovaSSumElemMatrixMPI(const InType &in) { |
| 14 | SetTypeOfTask(GetStaticTypeOfTask()); | ||
| 15 | GetInput() = in; | ||
| 16 | 20 | GetOutput() = 0.0; | |
| 17 | 20 | } | |
| 18 | |||
| 19 | 20 | bool FrolovaSSumElemMatrixMPI::ValidationImpl() { | |
| 20 | const auto &in = GetInput(); | ||
| 21 | |||
| 22 | const auto &my_matrix = std::get<0>(in); | ||
| 23 | 20 | int param_dim1 = std::get<1>(in); | |
| 24 | 20 | int param_dim2 = std::get<2>(in); | |
| 25 | |||
| 26 |
2/4✓ Branch 0 taken 20 times.
✗ Branch 1 not taken.
✗ Branch 2 not taken.
✓ Branch 3 taken 20 times.
|
20 | return (param_dim1 > 0 && param_dim2 > 0 && static_cast<int>(my_matrix.size()) == (param_dim1 * param_dim2)); |
| 27 | } | ||
| 28 | |||
| 29 | 20 | bool FrolovaSSumElemMatrixMPI::PreProcessingImpl() { | |
| 30 | 20 | return true; | |
| 31 | } | ||
| 32 | |||
| 33 | 20 | bool FrolovaSSumElemMatrixMPI::RunImpl() { | |
| 34 | 20 | int process_n = 0; | |
| 35 | 20 | MPI_Comm_size(MPI_COMM_WORLD, &process_n); | |
| 36 | |||
| 37 | 20 | int proc_rank = 0; | |
| 38 | 20 | MPI_Comm_rank(MPI_COMM_WORLD, &proc_rank); | |
| 39 | |||
| 40 | 20 | int total_size = 0; | |
| 41 | 20 | std::vector<double> vect_data; | |
| 42 | |||
| 43 |
2/2✓ Branch 0 taken 10 times.
✓ Branch 1 taken 10 times.
|
20 | if (proc_rank == 0) { |
| 44 |
1/2✓ Branch 1 taken 10 times.
✗ Branch 2 not taken.
|
10 | vect_data = std::get<0>(GetInput()); |
| 45 | 10 | total_size = static_cast<int>(vect_data.size()); | |
| 46 | |||
| 47 |
1/4✗ Branch 0 not taken.
✓ Branch 1 taken 10 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
|
10 | if (process_n > total_size && total_size > 0) { |
| 48 | ✗ | process_n = total_size; | |
| 49 | } | ||
| 50 | } | ||
| 51 | |||
| 52 |
1/2✓ Branch 1 taken 20 times.
✗ Branch 2 not taken.
|
20 | MPI_Bcast(&process_n, 1, MPI_INT, 0, MPI_COMM_WORLD); |
| 53 |
1/2✓ Branch 1 taken 20 times.
✗ Branch 2 not taken.
|
20 | MPI_Bcast(&total_size, 1, MPI_INT, 0, MPI_COMM_WORLD); |
| 54 | |||
| 55 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 20 times.
|
20 | if (total_size == 0) { |
| 56 | ✗ | if (proc_rank == 0) { | |
| 57 | ✗ | GetOutput() = 0.0; | |
| 58 | } | ||
| 59 | ✗ | return true; | |
| 60 | } | ||
| 61 | |||
| 62 |
2/2✓ Branch 0 taken 10 times.
✓ Branch 1 taken 10 times.
|
20 | if (proc_rank == 0) { |
| 63 |
1/2✓ Branch 1 taken 10 times.
✗ Branch 2 not taken.
|
10 | GetOutput() = ProcessMaster(process_n, vect_data); |
| 64 |
1/2✓ Branch 0 taken 10 times.
✗ Branch 1 not taken.
|
10 | } else if (proc_rank < process_n) { |
| 65 |
1/2✓ Branch 1 taken 10 times.
✗ Branch 2 not taken.
|
10 | ProcessWorker(); |
| 66 | 10 | GetOutput() = std::numeric_limits<double>::max(); | |
| 67 | } else { | ||
| 68 | ✗ | GetOutput() = std::numeric_limits<double>::max(); | |
| 69 | } | ||
| 70 | |||
| 71 | return true; | ||
| 72 | } | ||
| 73 | |||
| 74 | 20 | bool FrolovaSSumElemMatrixMPI::PostProcessingImpl() { | |
| 75 | 20 | return true; | |
| 76 | } | ||
| 77 | |||
| 78 | 10 | double FrolovaSSumElemMatrixMPI::ProcessMaster(int process_n, const std::vector<double> &vect_data) { | |
| 79 | 10 | const int n = static_cast<int>(vect_data.size()); | |
| 80 | 10 | const int base = n / process_n; | |
| 81 | 10 | int remain = n % process_n; | |
| 82 | |||
| 83 | int start_id = 0; | ||
| 84 | |||
| 85 |
2/2✓ Branch 0 taken 10 times.
✓ Branch 1 taken 10 times.
|
20 | for (int worker = 1; worker < process_n; worker++) { |
| 86 |
2/2✓ Branch 0 taken 5 times.
✓ Branch 1 taken 5 times.
|
10 | int part_size = base + (remain > 0 ? 1 : 0); |
| 87 |
2/2✓ Branch 0 taken 5 times.
✓ Branch 1 taken 5 times.
|
10 | if (remain > 0) { |
| 88 | 5 | remain--; | |
| 89 | } | ||
| 90 | |||
| 91 | 10 | MPI_Send(&part_size, 1, MPI_INT, worker, 0, MPI_COMM_WORLD); | |
| 92 | 10 | MPI_Send(vect_data.data() + start_id, part_size, MPI_DOUBLE, worker, 0, MPI_COMM_WORLD); | |
| 93 | |||
| 94 | 10 | start_id += part_size; | |
| 95 | } | ||
| 96 | |||
| 97 | double total_sum = 0.0; | ||
| 98 |
2/2✓ Branch 0 taken 2083 times.
✓ Branch 1 taken 10 times.
|
2093 | for (int i = start_id; i < n; ++i) { |
| 99 | 2083 | total_sum += vect_data[i]; | |
| 100 | } | ||
| 101 | |||
| 102 |
2/2✓ Branch 0 taken 10 times.
✓ Branch 1 taken 10 times.
|
20 | for (int worker = 1; worker < process_n; worker++) { |
| 103 | 10 | double worker_sum = 0.0; | |
| 104 | 10 | MPI_Recv(&worker_sum, 1, MPI_DOUBLE, worker, 0, MPI_COMM_WORLD, MPI_STATUS_IGNORE); | |
| 105 | 10 | total_sum += worker_sum; | |
| 106 | } | ||
| 107 | |||
| 108 | 10 | return total_sum; | |
| 109 | } | ||
| 110 | |||
| 111 | 10 | void FrolovaSSumElemMatrixMPI::ProcessWorker() { | |
| 112 | 10 | int part_size = 0; | |
| 113 | 10 | MPI_Recv(&part_size, 1, MPI_INT, 0, 0, MPI_COMM_WORLD, MPI_STATUS_IGNORE); | |
| 114 | |||
| 115 | 10 | std::vector<double> local_data(part_size); | |
| 116 |
1/2✓ Branch 1 taken 10 times.
✗ Branch 2 not taken.
|
10 | MPI_Recv(local_data.data(), part_size, MPI_DOUBLE, 0, 0, MPI_COMM_WORLD, MPI_STATUS_IGNORE); |
| 117 | |||
| 118 | 10 | double local_sum = 0.0; | |
| 119 |
2/2✓ Branch 0 taken 2088 times.
✓ Branch 1 taken 10 times.
|
2098 | for (double val : local_data) { |
| 120 | 2088 | local_sum += val; | |
| 121 | } | ||
| 122 | |||
| 123 |
1/2✓ Branch 1 taken 10 times.
✗ Branch 2 not taken.
|
10 | MPI_Send(&local_sum, 1, MPI_DOUBLE, 0, 0, MPI_COMM_WORLD); |
| 124 | 10 | } | |
| 125 | |||
| 126 | } // namespace frolova_s_sum_elem_matrix | ||
| 127 |