| Line | Branch | Exec | Source |
|---|---|---|---|
| 1 | #include "barkalova_m_mult_matrix_ccs/all/include/ops_all.hpp" | ||
| 2 | |||
| 3 | #include <mpi.h> | ||
| 4 | |||
| 5 | #include <algorithm> | ||
| 6 | #include <array> | ||
| 7 | #include <cmath> | ||
| 8 | #include <complex> | ||
| 9 | #include <cstddef> | ||
| 10 | #include <exception> | ||
| 11 | #include <utility> | ||
| 12 | #include <vector> | ||
| 13 | |||
| 14 | #include "barkalova_m_mult_matrix_ccs/common/include/common.hpp" | ||
| 15 | |||
| 16 | namespace barkalova_m_mult_matrix_ccs { | ||
| 17 | |||
| 18 |
1/2✓ Branch 2 taken 12 times.
✗ Branch 3 not taken.
|
12 | BarkalovaMMultMatrixCcsALL::BarkalovaMMultMatrixCcsALL(const InType &in) { |
| 19 | SetTypeOfTask(GetStaticTypeOfTask()); | ||
| 20 | |||
| 21 | 12 | int rank = 0; | |
| 22 |
1/2✓ Branch 1 taken 12 times.
✗ Branch 2 not taken.
|
12 | MPI_Comm_rank(MPI_COMM_WORLD, &rank); |
| 23 | |||
| 24 |
2/2✓ Branch 0 taken 6 times.
✓ Branch 1 taken 6 times.
|
12 | if (rank == 0) { |
| 25 | GetInput() = in; | ||
| 26 | } | ||
| 27 | |||
| 28 | 12 | GetOutput() = CCSMatrix{}; | |
| 29 | 12 | } | |
| 30 | |||
| 31 | 12 | bool BarkalovaMMultMatrixCcsALL::ValidationImpl() { | |
| 32 | 12 | int rank = 0; | |
| 33 | 12 | MPI_Comm_rank(MPI_COMM_WORLD, &rank); | |
| 34 | |||
| 35 |
2/2✓ Branch 0 taken 6 times.
✓ Branch 1 taken 6 times.
|
12 | if (rank != 0) { |
| 36 | return true; | ||
| 37 | } | ||
| 38 | |||
| 39 | const auto &[A, B] = GetInput(); | ||
| 40 |
1/2✓ Branch 0 taken 6 times.
✗ Branch 1 not taken.
|
6 | if (A.cols != B.rows) { |
| 41 | return false; | ||
| 42 | } | ||
| 43 |
3/6✓ Branch 0 taken 6 times.
✗ Branch 1 not taken.
✓ Branch 2 taken 6 times.
✗ Branch 3 not taken.
✓ Branch 4 taken 6 times.
✗ Branch 5 not taken.
|
6 | if (A.rows <= 0 || A.cols <= 0 || B.rows <= 0 || B.cols <= 0) { |
| 44 | return false; | ||
| 45 | } | ||
| 46 |
2/4✓ Branch 0 taken 6 times.
✗ Branch 1 not taken.
✓ Branch 2 taken 6 times.
✗ Branch 3 not taken.
|
6 | if (A.col_ptrs.size() != static_cast<size_t>(A.cols) + 1 || B.col_ptrs.size() != static_cast<size_t>(B.cols) + 1) { |
| 47 | return false; | ||
| 48 | } | ||
| 49 |
4/8✓ Branch 0 taken 6 times.
✗ Branch 1 not taken.
✓ Branch 2 taken 6 times.
✗ Branch 3 not taken.
✓ Branch 4 taken 6 times.
✗ Branch 5 not taken.
✓ Branch 6 taken 6 times.
✗ Branch 7 not taken.
|
6 | if (A.col_ptrs.empty() || A.col_ptrs[0] != 0 || B.col_ptrs.empty() || B.col_ptrs[0] != 0) { |
| 50 | return false; | ||
| 51 | } | ||
| 52 |
2/4✓ Branch 0 taken 6 times.
✗ Branch 1 not taken.
✓ Branch 2 taken 6 times.
✗ Branch 3 not taken.
|
12 | if (std::cmp_not_equal(A.nnz, A.values.size()) || std::cmp_not_equal(B.nnz, B.values.size())) { |
| 53 | ✗ | return false; | |
| 54 | } | ||
| 55 | return true; | ||
| 56 | } | ||
| 57 | |||
| 58 | 12 | bool BarkalovaMMultMatrixCcsALL::PreProcessingImpl() { | |
| 59 | 12 | return true; | |
| 60 | } | ||
| 61 | |||
| 62 | namespace { | ||
| 63 | constexpr double kEpsilon = 1e-10; | ||
| 64 | |||
| 65 | 12 | void TransponirMatr(const CCSMatrix &a, CCSMatrix &at) { | |
| 66 | 12 | at.rows = a.cols; | |
| 67 | 12 | at.cols = a.rows; | |
| 68 | 12 | at.nnz = a.nnz; | |
| 69 | |||
| 70 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 12 times.
|
12 | if (a.nnz == 0) { |
| 71 | at.values.clear(); | ||
| 72 | at.row_indices.clear(); | ||
| 73 | ✗ | at.col_ptrs.assign(at.cols + 1, 0); | |
| 74 | ✗ | return; | |
| 75 | } | ||
| 76 | |||
| 77 | 12 | std::vector<int> row_count(at.cols, 0); | |
| 78 |
2/2✓ Branch 0 taken 34 times.
✓ Branch 1 taken 12 times.
|
46 | for (int i = 0; i < a.nnz; i++) { |
| 79 | 34 | row_count[a.row_indices[i]]++; | |
| 80 | } | ||
| 81 | |||
| 82 |
1/2✓ Branch 1 taken 12 times.
✗ Branch 2 not taken.
|
12 | at.col_ptrs.resize(at.cols + 1); |
| 83 | 12 | at.col_ptrs[0] = 0; | |
| 84 |
2/2✓ Branch 0 taken 30 times.
✓ Branch 1 taken 12 times.
|
42 | for (int i = 0; i < at.cols; i++) { |
| 85 | 30 | at.col_ptrs[i + 1] = at.col_ptrs[i] + row_count[i]; | |
| 86 | } | ||
| 87 | |||
| 88 |
1/2✓ Branch 1 taken 12 times.
✗ Branch 2 not taken.
|
12 | at.values.resize(a.nnz); |
| 89 |
1/2✓ Branch 1 taken 12 times.
✗ Branch 2 not taken.
|
12 | at.row_indices.resize(a.nnz); |
| 90 | |||
| 91 |
1/4✓ Branch 1 taken 12 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
|
12 | std::vector<int> current_pos(at.cols, 0); |
| 92 |
2/2✓ Branch 0 taken 28 times.
✓ Branch 1 taken 12 times.
|
40 | for (int col = 0; col < a.cols; col++) { |
| 93 |
2/2✓ Branch 0 taken 34 times.
✓ Branch 1 taken 28 times.
|
62 | for (int i = a.col_ptrs[col]; i < a.col_ptrs[col + 1]; i++) { |
| 94 | 34 | int row = a.row_indices[i]; | |
| 95 | 34 | Complex val = a.values[i]; | |
| 96 | 34 | int pos = at.col_ptrs[row] + current_pos[row]; | |
| 97 | 34 | at.values[pos] = val; | |
| 98 | 34 | at.row_indices[pos] = col; | |
| 99 | 34 | current_pos[row]++; | |
| 100 | } | ||
| 101 | } | ||
| 102 | } | ||
| 103 | |||
| 104 | 24 | void BroadcastMatrix(CCSMatrix &matrix) { | |
| 105 | 24 | int rank = 0; | |
| 106 | 24 | MPI_Comm_rank(MPI_COMM_WORLD, &rank); | |
| 107 | |||
| 108 | 24 | std::array<int, 4> meta = {matrix.rows, matrix.cols, matrix.nnz, 0}; | |
| 109 |
2/2✓ Branch 0 taken 12 times.
✓ Branch 1 taken 12 times.
|
24 | if (rank == 0) { |
| 110 | 12 | meta[3] = static_cast<int>(matrix.col_ptrs.size()); | |
| 111 | } | ||
| 112 | 24 | MPI_Bcast(meta.data(), 4, MPI_INT, 0, MPI_COMM_WORLD); | |
| 113 | |||
| 114 | 24 | matrix.rows = meta[0]; | |
| 115 | 24 | matrix.cols = meta[1]; | |
| 116 | 24 | matrix.nnz = meta[2]; | |
| 117 | 24 | int col_ptrs_size = meta[3]; | |
| 118 | |||
| 119 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 24 times.
|
24 | if (matrix.nnz == 0) { |
| 120 | matrix.values.clear(); | ||
| 121 | matrix.row_indices.clear(); | ||
| 122 | ✗ | matrix.col_ptrs.assign(col_ptrs_size, 0); | |
| 123 | ✗ | return; | |
| 124 | } | ||
| 125 | |||
| 126 |
2/2✓ Branch 0 taken 12 times.
✓ Branch 1 taken 12 times.
|
24 | if (rank != 0) { |
| 127 | 12 | matrix.values.resize(matrix.nnz); | |
| 128 | 12 | matrix.row_indices.resize(matrix.nnz); | |
| 129 | 12 | matrix.col_ptrs.resize(col_ptrs_size); | |
| 130 | } | ||
| 131 | |||
| 132 | 24 | MPI_Bcast(matrix.col_ptrs.data(), col_ptrs_size, MPI_INT, 0, MPI_COMM_WORLD); | |
| 133 | 24 | MPI_Bcast(matrix.row_indices.data(), matrix.nnz, MPI_INT, 0, MPI_COMM_WORLD); | |
| 134 | |||
| 135 | 24 | std::vector<double> values_real(matrix.nnz); | |
| 136 |
1/4✓ Branch 1 taken 24 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
|
24 | std::vector<double> values_imag(matrix.nnz); |
| 137 | |||
| 138 |
2/2✓ Branch 0 taken 12 times.
✓ Branch 1 taken 12 times.
|
24 | if (rank == 0) { |
| 139 |
2/2✓ Branch 0 taken 42 times.
✓ Branch 1 taken 12 times.
|
54 | for (int i = 0; i < matrix.nnz; ++i) { |
| 140 | 42 | values_real[i] = matrix.values[i].real(); | |
| 141 | 42 | values_imag[i] = matrix.values[i].imag(); | |
| 142 | } | ||
| 143 | } | ||
| 144 | |||
| 145 |
1/2✓ Branch 1 taken 24 times.
✗ Branch 2 not taken.
|
24 | MPI_Bcast(values_real.data(), matrix.nnz, MPI_DOUBLE, 0, MPI_COMM_WORLD); |
| 146 |
1/2✓ Branch 1 taken 24 times.
✗ Branch 2 not taken.
|
24 | MPI_Bcast(values_imag.data(), matrix.nnz, MPI_DOUBLE, 0, MPI_COMM_WORLD); |
| 147 | |||
| 148 |
2/2✓ Branch 0 taken 12 times.
✓ Branch 1 taken 12 times.
|
24 | if (rank != 0) { |
| 149 |
2/2✓ Branch 0 taken 42 times.
✓ Branch 1 taken 12 times.
|
54 | for (int i = 0; i < matrix.nnz; ++i) { |
| 150 | 42 | matrix.values[i] = Complex(values_real[i], values_imag[i]); | |
| 151 | } | ||
| 152 | } | ||
| 153 | } | ||
| 154 | |||
| 155 | bool IsNonZero(const Complex &val) { | ||
| 156 | return std::abs(val.real()) > kEpsilon || std::abs(val.imag()) > kEpsilon; | ||
| 157 | } | ||
| 158 | |||
| 159 | 44 | Complex ComputeScalarProduct(const CCSMatrix &at, const CCSMatrix &b, int row_a, int col_b) { | |
| 160 | Complex sum = Complex(0.0, 0.0); | ||
| 161 | |||
| 162 | 44 | int ks = at.col_ptrs[row_a]; | |
| 163 | 44 | int ls = b.col_ptrs[col_b]; | |
| 164 | 44 | int kf = at.col_ptrs[row_a + 1]; | |
| 165 | 44 | int lf = b.col_ptrs[col_b + 1]; | |
| 166 | |||
| 167 |
2/2✓ Branch 0 taken 59 times.
✓ Branch 1 taken 44 times.
|
103 | while ((ks < kf) && (ls < lf)) { |
| 168 |
2/2✓ Branch 0 taken 11 times.
✓ Branch 1 taken 48 times.
|
59 | if (at.row_indices[ks] < b.row_indices[ls]) { |
| 169 | 11 | ks++; | |
| 170 |
2/2✓ Branch 0 taken 14 times.
✓ Branch 1 taken 34 times.
|
48 | } else if (at.row_indices[ks] > b.row_indices[ls]) { |
| 171 | 14 | ls++; | |
| 172 | } else { | ||
| 173 | sum += at.values[ks] * b.values[ls]; | ||
| 174 | 34 | ks++; | |
| 175 | 34 | ls++; | |
| 176 | } | ||
| 177 | } | ||
| 178 | |||
| 179 | 44 | return sum; | |
| 180 | } | ||
| 181 | |||
| 182 | void ComputeLocalColumns(int start_col, int local_cols, const CCSMatrix &at, const CCSMatrix &b, | ||
| 183 | std::vector<std::vector<int>> &col_rows, std::vector<std::vector<Complex>> &col_vals) { | ||
| 184 | 12 | #pragma omp parallel for schedule(static) default(none) shared(start_col, local_cols, at, b, col_rows, col_vals) | |
| 185 | for (int j_local = 0; j_local < local_cols; ++j_local) { | ||
| 186 | int global_col = start_col + j_local; | ||
| 187 | |||
| 188 | std::vector<int> rows; | ||
| 189 | std::vector<Complex> vals; | ||
| 190 | rows.reserve(100); | ||
| 191 | vals.reserve(100); | ||
| 192 | |||
| 193 | for (int i = 0; i < at.cols; i++) { | ||
| 194 | Complex sum = ComputeScalarProduct(at, b, i, global_col); | ||
| 195 | if (IsNonZero(sum)) { | ||
| 196 | rows.push_back(i); | ||
| 197 | vals.push_back(sum); | ||
| 198 | } | ||
| 199 | } | ||
| 200 | |||
| 201 | col_rows[j_local] = std::move(rows); | ||
| 202 | col_vals[j_local] = std::move(vals); | ||
| 203 | } | ||
| 204 | } | ||
| 205 | |||
| 206 | 12 | void LocVectors(int local_cols, const std::vector<std::vector<int>> &col_rows, | |
| 207 | const std::vector<std::vector<Complex>> &col_vals, std::vector<int> &local_row_indices, | ||
| 208 | std::vector<Complex> &local_values) { | ||
| 209 |
2/2✓ Branch 0 taken 17 times.
✓ Branch 1 taken 12 times.
|
29 | for (int j = 0; j < local_cols; ++j) { |
| 210 | 17 | local_row_indices.insert(local_row_indices.end(), col_rows[j].begin(), col_rows[j].end()); | |
| 211 | 17 | local_values.insert(local_values.end(), col_vals[j].begin(), col_vals[j].end()); | |
| 212 | } | ||
| 213 | 12 | } | |
| 214 | |||
| 215 | 12 | void GatherResults(int rank, int size, int local_nnz, const std::vector<int> &local_row_indices, | |
| 216 | const std::vector<Complex> &local_values, std::vector<int> &global_row_indices, | ||
| 217 | std::vector<double> &global_values_real, std::vector<double> &global_values_imag, int total_nnz) { | ||
| 218 |
1/2✓ Branch 2 taken 12 times.
✗ Branch 3 not taken.
|
12 | std::vector<int> recv_counts(size, 0); |
| 219 |
1/2✓ Branch 1 taken 12 times.
✗ Branch 2 not taken.
|
12 | MPI_Gather(&local_nnz, 1, MPI_INT, recv_counts.data(), 1, MPI_INT, 0, MPI_COMM_WORLD); |
| 220 | |||
| 221 |
1/4✓ Branch 1 taken 12 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
|
12 | std::vector<int> displs(size, 0); |
| 222 |
2/2✓ Branch 0 taken 6 times.
✓ Branch 1 taken 6 times.
|
12 | if (rank == 0) { |
| 223 |
2/2✓ Branch 0 taken 6 times.
✓ Branch 1 taken 6 times.
|
12 | for (int i = 1; i < size; ++i) { |
| 224 | 6 | displs[i] = displs[i - 1] + recv_counts[i - 1]; | |
| 225 | } | ||
| 226 | } | ||
| 227 | |||
| 228 | if (rank == 0) { | ||
| 229 |
1/2✓ Branch 1 taken 6 times.
✗ Branch 2 not taken.
|
6 | global_row_indices.resize(total_nnz); |
| 230 | } | ||
| 231 |
1/2✓ Branch 1 taken 12 times.
✗ Branch 2 not taken.
|
12 | MPI_Gatherv(local_row_indices.data(), local_nnz, MPI_INT, global_row_indices.data(), recv_counts.data(), |
| 232 | displs.data(), MPI_INT, 0, MPI_COMM_WORLD); | ||
| 233 | |||
| 234 |
1/4✓ Branch 1 taken 12 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
|
12 | std::vector<double> local_values_real(local_nnz); |
| 235 |
1/4✓ Branch 1 taken 12 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
|
12 | std::vector<double> local_values_imag(local_nnz); |
| 236 |
2/2✓ Branch 0 taken 29 times.
✓ Branch 1 taken 12 times.
|
41 | for (int i = 0; i < local_nnz; ++i) { |
| 237 | 29 | local_values_real[i] = local_values[i].real(); | |
| 238 | 29 | local_values_imag[i] = local_values[i].imag(); | |
| 239 | } | ||
| 240 | |||
| 241 |
2/2✓ Branch 0 taken 6 times.
✓ Branch 1 taken 6 times.
|
12 | if (rank == 0) { |
| 242 |
1/2✓ Branch 1 taken 6 times.
✗ Branch 2 not taken.
|
6 | global_values_real.resize(total_nnz); |
| 243 | } | ||
| 244 |
1/2✓ Branch 1 taken 12 times.
✗ Branch 2 not taken.
|
12 | MPI_Gatherv(local_values_real.data(), local_nnz, MPI_DOUBLE, global_values_real.data(), recv_counts.data(), |
| 245 | displs.data(), MPI_DOUBLE, 0, MPI_COMM_WORLD); | ||
| 246 | |||
| 247 |
2/2✓ Branch 0 taken 6 times.
✓ Branch 1 taken 6 times.
|
12 | if (rank == 0) { |
| 248 |
1/2✓ Branch 1 taken 6 times.
✗ Branch 2 not taken.
|
6 | global_values_imag.resize(total_nnz); |
| 249 | } | ||
| 250 |
1/2✓ Branch 1 taken 12 times.
✗ Branch 2 not taken.
|
12 | MPI_Gatherv(local_values_imag.data(), local_nnz, MPI_DOUBLE, global_values_imag.data(), recv_counts.data(), |
| 251 | displs.data(), MPI_DOUBLE, 0, MPI_COMM_WORLD); | ||
| 252 | 12 | } | |
| 253 | |||
| 254 | 12 | void BroadcastRes(int rank, int total_rows, int total_cols, int total_nnz, std::vector<int> &global_col_ptrs, | |
| 255 | std::vector<int> &global_row_indices, std::vector<double> &global_values_real, | ||
| 256 | std::vector<double> &global_values_imag) { | ||
| 257 | 12 | int bcast_rows = total_rows; | |
| 258 | 12 | int bcast_cols = total_cols; | |
| 259 | 12 | int bcast_nnz = total_nnz; | |
| 260 | |||
| 261 | 12 | MPI_Bcast(&bcast_rows, 1, MPI_INT, 0, MPI_COMM_WORLD); | |
| 262 | 12 | MPI_Bcast(&bcast_cols, 1, MPI_INT, 0, MPI_COMM_WORLD); | |
| 263 | 12 | MPI_Bcast(&bcast_nnz, 1, MPI_INT, 0, MPI_COMM_WORLD); | |
| 264 | |||
| 265 | 12 | int col_ptrs_size = static_cast<int>(global_col_ptrs.size()); | |
| 266 | 12 | MPI_Bcast(&col_ptrs_size, 1, MPI_INT, 0, MPI_COMM_WORLD); | |
| 267 |
2/2✓ Branch 0 taken 6 times.
✓ Branch 1 taken 6 times.
|
12 | if (rank != 0) { |
| 268 | 6 | global_col_ptrs.resize(col_ptrs_size); | |
| 269 | } | ||
| 270 | 12 | MPI_Bcast(global_col_ptrs.data(), col_ptrs_size, MPI_INT, 0, MPI_COMM_WORLD); | |
| 271 | |||
| 272 |
2/2✓ Branch 0 taken 6 times.
✓ Branch 1 taken 6 times.
|
12 | if (rank != 0) { |
| 273 | 6 | global_row_indices.resize(bcast_nnz); | |
| 274 | } | ||
| 275 | 12 | MPI_Bcast(global_row_indices.data(), bcast_nnz, MPI_INT, 0, MPI_COMM_WORLD); | |
| 276 | |||
| 277 |
2/2✓ Branch 0 taken 6 times.
✓ Branch 1 taken 6 times.
|
12 | if (rank != 0) { |
| 278 | 6 | global_values_real.resize(bcast_nnz); | |
| 279 | 6 | global_values_imag.resize(bcast_nnz); | |
| 280 | } | ||
| 281 | 12 | MPI_Bcast(global_values_real.data(), bcast_nnz, MPI_DOUBLE, 0, MPI_COMM_WORLD); | |
| 282 | 12 | MPI_Bcast(global_values_imag.data(), bcast_nnz, MPI_DOUBLE, 0, MPI_COMM_WORLD); | |
| 283 | 12 | } | |
| 284 | |||
| 285 | } // namespace | ||
| 286 | |||
| 287 | 12 | bool BarkalovaMMultMatrixCcsALL::RunImpl() { | |
| 288 | try { | ||
| 289 | 12 | int rank = 0; | |
| 290 | 12 | int size = 0; | |
| 291 |
1/2✓ Branch 1 taken 12 times.
✗ Branch 2 not taken.
|
12 | MPI_Comm_rank(MPI_COMM_WORLD, &rank); |
| 292 |
1/2✓ Branch 1 taken 12 times.
✗ Branch 2 not taken.
|
12 | MPI_Comm_size(MPI_COMM_WORLD, &size); |
| 293 | |||
| 294 |
1/2✓ Branch 1 taken 12 times.
✗ Branch 2 not taken.
|
12 | CCSMatrix a = GetInput().first; |
| 295 |
1/2✓ Branch 1 taken 12 times.
✗ Branch 2 not taken.
|
12 | CCSMatrix b = GetInput().second; |
| 296 | |||
| 297 |
1/2✓ Branch 1 taken 12 times.
✗ Branch 2 not taken.
|
12 | BroadcastMatrix(a); |
| 298 |
1/2✓ Branch 1 taken 12 times.
✗ Branch 2 not taken.
|
12 | BroadcastMatrix(b); |
| 299 | |||
| 300 | 12 | CCSMatrix at; | |
| 301 |
1/2✓ Branch 1 taken 12 times.
✗ Branch 2 not taken.
|
12 | TransponirMatr(a, at); |
| 302 | |||
| 303 | 12 | const int total_cols = b.cols; | |
| 304 | 12 | const int total_rows = a.rows; | |
| 305 | |||
| 306 | 12 | int cols_per_process = total_cols / size; | |
| 307 | 12 | int remainder = total_cols % size; | |
| 308 | |||
| 309 |
2/2✓ Branch 0 taken 9 times.
✓ Branch 1 taken 3 times.
|
12 | int start_col = (rank * cols_per_process) + std::min(rank, remainder); |
| 310 |
2/2✓ Branch 0 taken 9 times.
✓ Branch 1 taken 3 times.
|
12 | int local_cols = cols_per_process + (rank < remainder ? 1 : 0); |
| 311 | |||
| 312 |
1/2✓ Branch 1 taken 12 times.
✗ Branch 2 not taken.
|
12 | std::vector<std::vector<int>> col_rows(local_cols); |
| 313 |
1/2✓ Branch 1 taken 12 times.
✗ Branch 2 not taken.
|
12 | std::vector<std::vector<Complex>> col_vals(local_cols); |
| 314 | |||
| 315 | ComputeLocalColumns(start_col, local_cols, at, b, col_rows, col_vals); | ||
| 316 | |||
| 317 | 12 | std::vector<int> local_row_indices; | |
| 318 | 12 | std::vector<Complex> local_values; | |
| 319 |
1/2✓ Branch 1 taken 12 times.
✗ Branch 2 not taken.
|
12 | LocVectors(local_cols, col_rows, col_vals, local_row_indices, local_values); |
| 320 | 12 | int local_nnz = static_cast<int>(local_values.size()); | |
| 321 | |||
| 322 |
1/4✓ Branch 1 taken 12 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
|
12 | std::vector<int> global_col_nnz(total_cols, 0); |
| 323 |
2/2✓ Branch 0 taken 17 times.
✓ Branch 1 taken 12 times.
|
29 | for (int j = 0; j < local_cols; ++j) { |
| 324 | 17 | global_col_nnz[start_col + j] = static_cast<int>(col_rows[j].size()); | |
| 325 | } | ||
| 326 | |||
| 327 |
2/6✓ Branch 1 taken 12 times.
✗ Branch 2 not taken.
✓ Branch 4 taken 12 times.
✗ Branch 5 not taken.
✗ Branch 6 not taken.
✗ Branch 7 not taken.
|
12 | std::vector<int> recv_col_nnz(total_cols); |
| 328 |
1/2✓ Branch 1 taken 12 times.
✗ Branch 2 not taken.
|
12 | MPI_Allreduce(global_col_nnz.data(), recv_col_nnz.data(), total_cols, MPI_INT, MPI_SUM, MPI_COMM_WORLD); |
| 329 | |||
| 330 |
1/4✓ Branch 1 taken 12 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
|
12 | std::vector<int> global_col_ptrs = {0}; |
| 331 |
2/2✓ Branch 0 taken 34 times.
✓ Branch 1 taken 12 times.
|
46 | for (int j = 0; j < total_cols; ++j) { |
| 332 |
1/4✓ Branch 1 taken 34 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
|
34 | global_col_ptrs.push_back(global_col_ptrs.back() + recv_col_nnz[j]); |
| 333 | } | ||
| 334 | 12 | int total_nnz = global_col_ptrs.back(); | |
| 335 | |||
| 336 | 12 | std::vector<int> global_row_indices; | |
| 337 | 12 | std::vector<double> global_values_real; | |
| 338 | 12 | std::vector<double> global_values_imag; | |
| 339 |
1/2✓ Branch 1 taken 12 times.
✗ Branch 2 not taken.
|
12 | GatherResults(rank, size, local_nnz, local_row_indices, local_values, global_row_indices, global_values_real, |
| 340 | global_values_imag, total_nnz); | ||
| 341 | |||
| 342 |
1/2✓ Branch 1 taken 12 times.
✗ Branch 2 not taken.
|
12 | BroadcastRes(rank, total_rows, total_cols, total_nnz, global_col_ptrs, global_row_indices, global_values_real, |
| 343 | global_values_imag); | ||
| 344 | |||
| 345 |
1/4✓ Branch 1 taken 12 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
|
12 | std::vector<Complex> global_values(total_nnz); |
| 346 |
2/2✓ Branch 0 taken 58 times.
✓ Branch 1 taken 12 times.
|
70 | for (int i = 0; i < total_nnz; ++i) { |
| 347 | 58 | global_values[i] = Complex(global_values_real[i], global_values_imag[i]); | |
| 348 | } | ||
| 349 | |||
| 350 | 12 | CCSMatrix c; | |
| 351 | 12 | c.rows = total_rows; | |
| 352 | 12 | c.cols = total_cols; | |
| 353 | 12 | c.nnz = total_nnz; | |
| 354 | c.values = std::move(global_values); | ||
| 355 | c.row_indices = std::move(global_row_indices); | ||
| 356 | c.col_ptrs = std::move(global_col_ptrs); | ||
| 357 | |||
| 358 |
1/2✓ Branch 1 taken 12 times.
✗ Branch 2 not taken.
|
12 | GetOutput() = c; |
| 359 | return true; | ||
| 360 | |||
| 361 |
0/2✗ Branch 12 not taken.
✗ Branch 13 not taken.
|
24 | } catch (const std::exception &) { |
| 362 | return false; | ||
| 363 | ✗ | } | |
| 364 | } | ||
| 365 | |||
| 366 | 12 | bool BarkalovaMMultMatrixCcsALL::PostProcessingImpl() { | |
| 367 | 12 | int rank = 0; | |
| 368 | 12 | MPI_Comm_rank(MPI_COMM_WORLD, &rank); | |
| 369 | |||
| 370 |
2/2✓ Branch 0 taken 6 times.
✓ Branch 1 taken 6 times.
|
12 | if (rank != 0) { |
| 371 | return true; | ||
| 372 | } | ||
| 373 | |||
| 374 | const auto &c = GetOutput(); | ||
| 375 |
2/4✓ Branch 0 taken 6 times.
✗ Branch 1 not taken.
✗ Branch 2 not taken.
✓ Branch 3 taken 6 times.
|
6 | if (c.rows <= 0 || c.cols <= 0) { |
| 376 | return false; | ||
| 377 | } | ||
| 378 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 6 times.
|
6 | if (c.col_ptrs.size() != static_cast<size_t>(c.cols) + 1) { |
| 379 | return false; | ||
| 380 | } | ||
| 381 |
2/2✓ Branch 0 taken 17 times.
✓ Branch 1 taken 6 times.
|
23 | for (size_t i = 1; i < c.col_ptrs.size(); ++i) { |
| 382 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 17 times.
|
17 | if (c.col_ptrs[i] < c.col_ptrs[i - 1]) { |
| 383 | return false; | ||
| 384 | } | ||
| 385 | } | ||
| 386 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 6 times.
|
6 | if (std::cmp_not_equal(c.nnz, c.values.size()) || std::cmp_not_equal(c.nnz, c.row_indices.size())) { |
| 387 | return false; | ||
| 388 | } | ||
| 389 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 6 times.
|
6 | if (c.col_ptrs[0] != 0) { |
| 390 | return false; | ||
| 391 | } | ||
| 392 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 6 times.
|
6 | if (c.col_ptrs.back() != c.nnz) { |
| 393 | return false; | ||
| 394 | } | ||
| 395 | return true; | ||
| 396 | } | ||
| 397 | |||
| 398 | } // namespace barkalova_m_mult_matrix_ccs | ||
| 399 |