| Line | Branch | Exec | Source |
|---|---|---|---|
| 1 | #include "zenin_a_sum_values_by_columns_matrix/mpi/include/ops_mpi.hpp" | ||
| 2 | |||
| 3 | #include <mpi.h> | ||
| 4 | |||
| 5 | #include <cmath> | ||
| 6 | #include <cstddef> | ||
| 7 | #include <stdexcept> | ||
| 8 | #include <tuple> | ||
| 9 | #include <utility> | ||
| 10 | #include <vector> | ||
| 11 | |||
| 12 | #include "zenin_a_sum_values_by_columns_matrix/common/include/common.hpp" | ||
| 13 | |||
| 14 | namespace zenin_a_sum_values_by_columns_matrix { | ||
| 15 | |||
| 16 |
1/2✓ Branch 1 taken 30 times.
✗ Branch 2 not taken.
|
30 | ZeninASumValuesByColumnsMatrixMPI::ZeninASumValuesByColumnsMatrixMPI(const InType &in) { |
| 17 | SetTypeOfTask(GetStaticTypeOfTask()); | ||
| 18 | GetInput() = in; | ||
| 19 | 30 | GetOutput() = OutType{}; | |
| 20 | 30 | } | |
| 21 | |||
| 22 | 30 | bool ZeninASumValuesByColumnsMatrixMPI::ValidationImpl() { | |
| 23 | auto &input = GetInput(); | ||
| 24 |
2/4✓ Branch 0 taken 30 times.
✗ Branch 1 not taken.
✗ Branch 2 not taken.
✓ Branch 3 taken 30 times.
|
30 | return ((std::get<0>(input)) * std::get<1>(input) == std::get<2>(input).size() && (GetOutput().empty())); |
| 25 | } | ||
| 26 | |||
| 27 | 30 | bool ZeninASumValuesByColumnsMatrixMPI::PreProcessingImpl() { | |
| 28 | 30 | int rank = 0; | |
| 29 | 30 | MPI_Comm_rank(MPI_COMM_WORLD, &rank); | |
| 30 |
2/2✓ Branch 0 taken 15 times.
✓ Branch 1 taken 15 times.
|
30 | if (rank != 0) { |
| 31 | return true; | ||
| 32 | } | ||
| 33 | GetOutput().clear(); | ||
| 34 | return true; | ||
| 35 | } | ||
| 36 | |||
| 37 | 15 | void ZeninASumValuesByColumnsMatrixMPI::FillSendBuffer(const std::vector<double> &mat, std::vector<double> &sendbuf, | |
| 38 | size_t rows, size_t cols, size_t base, size_t rest, | ||
| 39 | int world_size) { | ||
| 40 | size_t pos = 0; | ||
| 41 |
2/2✓ Branch 0 taken 30 times.
✓ Branch 1 taken 15 times.
|
45 | for (int proc = 0; proc < world_size; proc++) { |
| 42 | 30 | auto proc_size = static_cast<size_t>(proc); | |
| 43 | 30 | size_t pc_begin = (proc_size * base) + (std::cmp_less(proc_size, rest) ? proc_size : rest); | |
| 44 |
2/2✓ Branch 0 taken 21 times.
✓ Branch 1 taken 9 times.
|
30 | size_t pc_end = pc_begin + (base + (std::cmp_less(proc_size, rest) ? 1 : 0)); |
| 45 |
2/2✓ Branch 0 taken 11697 times.
✓ Branch 1 taken 30 times.
|
11727 | for (size_t col = pc_begin; col < pc_end; col++) { |
| 46 |
2/2✓ Branch 0 taken 1061987 times.
✓ Branch 1 taken 11697 times.
|
1073684 | for (size_t row = 0; row < rows; row++) { |
| 47 | 1061987 | sendbuf[pos++] = mat[(row * cols) + col]; | |
| 48 | } | ||
| 49 | } | ||
| 50 | } | ||
| 51 | 15 | } | |
| 52 | |||
| 53 | 30 | bool ZeninASumValuesByColumnsMatrixMPI::RunImpl() { | |
| 54 | 30 | auto rows = static_cast<size_t>(std::get<0>(GetInput())); | |
| 55 | 30 | auto cols = static_cast<size_t>(std::get<1>(GetInput())); | |
| 56 | const std::vector<double> &mat = std::get<2>(GetInput()); | ||
| 57 | |||
| 58 | std::vector<double> &global_sum = GetOutput(); | ||
| 59 | |||
| 60 | 30 | int rank = 0; | |
| 61 | 30 | int world_size = 0; | |
| 62 | |||
| 63 | 30 | MPI_Comm_rank(MPI_COMM_WORLD, &rank); | |
| 64 | 30 | MPI_Comm_size(MPI_COMM_WORLD, &world_size); | |
| 65 | |||
| 66 | 30 | const size_t base = cols / static_cast<size_t>(world_size); | |
| 67 | 30 | const size_t rest = cols % static_cast<size_t>(world_size); | |
| 68 | |||
| 69 |
2/2✓ Branch 0 taken 21 times.
✓ Branch 1 taken 9 times.
|
30 | const size_t my_cols = base + (std::cmp_less(static_cast<size_t>(rank), rest) ? 1 : 0); |
| 70 | |||
| 71 | 30 | std::vector<int> sendcounts(static_cast<size_t>(world_size)); | |
| 72 |
1/4✓ Branch 1 taken 30 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
|
30 | std::vector<int> displs(static_cast<size_t>(world_size)); |
| 73 |
2/2✓ Branch 0 taken 15 times.
✓ Branch 1 taken 15 times.
|
30 | if (rank == 0) { |
| 74 | int offset = 0; | ||
| 75 |
2/2✓ Branch 0 taken 30 times.
✓ Branch 1 taken 15 times.
|
45 | for (int proc = 0; proc < world_size; proc++) { |
| 76 |
2/2✓ Branch 0 taken 21 times.
✓ Branch 1 taken 9 times.
|
30 | size_t pc = base + (std::cmp_less(static_cast<size_t>(proc), rest) ? 1 : 0); |
| 77 | 30 | sendcounts[proc] = static_cast<int>(pc * rows); | |
| 78 | 30 | displs[proc] = offset; | |
| 79 | 30 | offset += sendcounts[proc]; | |
| 80 | } | ||
| 81 | } | ||
| 82 | |||
| 83 | 30 | std::vector<double> sendbuf; | |
| 84 |
2/2✓ Branch 0 taken 15 times.
✓ Branch 1 taken 15 times.
|
30 | if (rank == 0) { |
| 85 |
1/2✓ Branch 1 taken 15 times.
✗ Branch 2 not taken.
|
15 | sendbuf.resize(rows * cols); |
| 86 | 15 | FillSendBuffer(mat, sendbuf, rows, cols, base, rest, world_size); | |
| 87 | } | ||
| 88 |
2/6✓ Branch 1 taken 30 times.
✗ Branch 2 not taken.
✓ Branch 4 taken 30 times.
✗ Branch 5 not taken.
✗ Branch 6 not taken.
✗ Branch 7 not taken.
|
30 | std::vector<double> local_block(rows * my_cols); |
| 89 |
1/2✓ Branch 1 taken 30 times.
✗ Branch 2 not taken.
|
30 | MPI_Scatterv(sendbuf.data(), sendcounts.data(), displs.data(), MPI_DOUBLE, local_block.data(), |
| 90 | static_cast<int>(local_block.size()), MPI_DOUBLE, 0, MPI_COMM_WORLD); | ||
| 91 | |||
| 92 |
1/4✓ Branch 1 taken 30 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
|
30 | std::vector<double> local_sum(my_cols, 0.0); |
| 93 |
2/2✓ Branch 0 taken 11697 times.
✓ Branch 1 taken 30 times.
|
11727 | for (size_t col_id = 0; col_id < my_cols; col_id++) { |
| 94 |
2/2✓ Branch 0 taken 1061987 times.
✓ Branch 1 taken 11697 times.
|
1073684 | for (size_t row_id = 0; row_id < rows; row_id++) { |
| 95 | 1061987 | local_sum[col_id] += local_block[(col_id * rows) + row_id]; | |
| 96 | } | ||
| 97 | } | ||
| 98 |
1/4✓ Branch 1 taken 30 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
|
30 | std::vector<int> recvcounts(static_cast<size_t>(world_size)); |
| 99 |
1/4✓ Branch 1 taken 30 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
|
30 | std::vector<int> recvdispls(static_cast<size_t>(world_size)); |
| 100 | |||
| 101 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 30 times.
|
30 | if (rows == 0) { |
| 102 | ✗ | throw std::runtime_error("Matrix has zero rows"); | |
| 103 | } | ||
| 104 | |||
| 105 |
2/2✓ Branch 0 taken 15 times.
✓ Branch 1 taken 15 times.
|
30 | if (rank == 0) { |
| 106 | size_t offset = 0; | ||
| 107 |
2/2✓ Branch 0 taken 30 times.
✓ Branch 1 taken 15 times.
|
45 | for (int proc = 0; proc < world_size; proc++) { |
| 108 | 30 | recvcounts[proc] = sendcounts[proc] / static_cast<int>(rows); | |
| 109 | 30 | recvdispls[proc] = static_cast<int>(offset); | |
| 110 | 30 | offset += static_cast<size_t>(recvcounts[proc]); | |
| 111 | } | ||
| 112 |
1/4✓ Branch 1 taken 15 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
|
15 | global_sum.assign(cols, 0.0); |
| 113 | } | ||
| 114 | |||
| 115 |
1/2✓ Branch 1 taken 30 times.
✗ Branch 2 not taken.
|
30 | MPI_Gatherv(local_sum.data(), static_cast<int>(my_cols), MPI_DOUBLE, global_sum.data(), recvcounts.data(), |
| 116 | recvdispls.data(), MPI_DOUBLE, 0, MPI_COMM_WORLD); | ||
| 117 |
1/2✓ Branch 1 taken 30 times.
✗ Branch 2 not taken.
|
30 | global_sum.resize(cols); |
| 118 |
1/2✓ Branch 1 taken 30 times.
✗ Branch 2 not taken.
|
30 | MPI_Bcast(global_sum.data(), static_cast<int>(cols), MPI_DOUBLE, 0, MPI_COMM_WORLD); |
| 119 | 30 | return true; | |
| 120 | } | ||
| 121 | |||
| 122 | 30 | bool ZeninASumValuesByColumnsMatrixMPI::PostProcessingImpl() { | |
| 123 | 30 | return true; | |
| 124 | } | ||
| 125 | |||
| 126 | } // namespace zenin_a_sum_values_by_columns_matrix | ||
| 127 |