| Line | Branch | Exec | Source |
|---|---|---|---|
| 1 | #include "kotelnikova_a_double_matr_mult/stl/include/ops_stl.hpp" | ||
| 2 | |||
| 3 | #include <algorithm> | ||
| 4 | #include <cmath> | ||
| 5 | #include <cstddef> | ||
| 6 | #include <thread> | ||
| 7 | #include <vector> | ||
| 8 | |||
| 9 | #include "kotelnikova_a_double_matr_mult/common/include/common.hpp" | ||
| 10 | |||
| 11 | namespace kotelnikova_a_double_matr_mult { | ||
| 12 | |||
| 13 |
1/2✓ Branch 2 taken 40 times.
✗ Branch 3 not taken.
|
40 | KotelnikovaATaskSTL::KotelnikovaATaskSTL(const InType &in) { |
| 14 | SetTypeOfTask(GetStaticTypeOfTask()); | ||
| 15 | GetInput() = in; | ||
| 16 | 40 | GetOutput() = SparseMatrixCCS(); | |
| 17 | 40 | } | |
| 18 | |||
| 19 | 80 | bool KotelnikovaATaskSTL::IsMatrixValid(const SparseMatrixCCS &matrix) { | |
| 20 |
2/4✓ Branch 0 taken 80 times.
✗ Branch 1 not taken.
✗ Branch 2 not taken.
✓ Branch 3 taken 80 times.
|
80 | if (matrix.rows < 0 || matrix.cols < 0) { |
| 21 | return false; | ||
| 22 | } | ||
| 23 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 80 times.
|
80 | if (matrix.col_ptrs.size() != static_cast<size_t>(matrix.cols) + 1) { |
| 24 | return false; | ||
| 25 | } | ||
| 26 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 80 times.
|
80 | if (matrix.values.size() != matrix.row_indices.size()) { |
| 27 | return false; | ||
| 28 | } | ||
| 29 | |||
| 30 |
2/4✗ Branch 0 not taken.
✓ Branch 1 taken 80 times.
✗ Branch 2 not taken.
✓ Branch 3 taken 80 times.
|
80 | if (matrix.col_ptrs.empty() || matrix.col_ptrs[0] != 0) { |
| 31 | return false; | ||
| 32 | } | ||
| 33 | |||
| 34 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 80 times.
|
80 | const int total_elements = static_cast<int>(matrix.values.size()); |
| 35 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 80 times.
|
80 | if (matrix.col_ptrs[matrix.cols] != total_elements) { |
| 36 | return false; | ||
| 37 | } | ||
| 38 | |||
| 39 |
2/2✓ Branch 0 taken 232 times.
✓ Branch 1 taken 80 times.
|
312 | for (size_t i = 0; i < matrix.col_ptrs.size() - 1; ++i) { |
| 40 |
2/4✗ Branch 0 not taken.
✓ Branch 1 taken 232 times.
✗ Branch 2 not taken.
✓ Branch 3 taken 232 times.
|
232 | if (matrix.col_ptrs[i] > matrix.col_ptrs[i + 1] || matrix.col_ptrs[i] < 0) { |
| 41 | return false; | ||
| 42 | } | ||
| 43 | } | ||
| 44 | |||
| 45 |
2/2✓ Branch 0 taken 288 times.
✓ Branch 1 taken 80 times.
|
368 | for (size_t i = 0; i < matrix.row_indices.size(); ++i) { |
| 46 |
2/4✗ Branch 0 not taken.
✓ Branch 1 taken 288 times.
✗ Branch 2 not taken.
✓ Branch 3 taken 288 times.
|
288 | if (matrix.row_indices[i] < 0 || matrix.row_indices[i] >= matrix.rows) { |
| 47 | return false; | ||
| 48 | } | ||
| 49 | } | ||
| 50 | |||
| 51 | return true; | ||
| 52 | } | ||
| 53 | |||
| 54 |
1/2✓ Branch 0 taken 40 times.
✗ Branch 1 not taken.
|
40 | bool KotelnikovaATaskSTL::ValidationImpl() { |
| 55 | const auto &[a, b] = GetInput(); | ||
| 56 | |||
| 57 |
2/4✓ Branch 0 taken 40 times.
✗ Branch 1 not taken.
✓ Branch 2 taken 40 times.
✗ Branch 3 not taken.
|
40 | if (!IsMatrixValid(a) || !IsMatrixValid(b)) { |
| 58 | return false; | ||
| 59 | } | ||
| 60 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 40 times.
|
40 | if (a.cols != b.rows) { |
| 61 | ✗ | return false; | |
| 62 | } | ||
| 63 | |||
| 64 | return true; | ||
| 65 | } | ||
| 66 | |||
| 67 | 40 | bool KotelnikovaATaskSTL::PreProcessingImpl() { | |
| 68 | const auto &[a, b] = GetInput(); | ||
| 69 | 40 | GetOutput() = SparseMatrixCCS(a.rows, b.cols); | |
| 70 | 40 | return true; | |
| 71 | } | ||
| 72 | |||
| 73 | namespace { | ||
| 74 | |||
| 75 | 224 | std::vector<double> ComputeColumn(const SparseMatrixCCS &a, const SparseMatrixCCS &b, int col_idx) { | |
| 76 | 224 | std::vector<double> temp(static_cast<size_t>(a.rows), 0.0); | |
| 77 | |||
| 78 |
2/2✓ Branch 0 taken 224 times.
✓ Branch 1 taken 224 times.
|
448 | for (int b_idx = b.col_ptrs[col_idx]; b_idx < b.col_ptrs[col_idx + 1]; ++b_idx) { |
| 79 | 224 | const int k = b.row_indices[b_idx]; | |
| 80 | 224 | const double b_val = b.values[b_idx]; | |
| 81 | |||
| 82 |
2/2✓ Branch 0 taken 336 times.
✓ Branch 1 taken 224 times.
|
560 | for (int a_idx = a.col_ptrs[k]; a_idx < a.col_ptrs[k + 1]; ++a_idx) { |
| 83 | 336 | const int i = a.row_indices[a_idx]; | |
| 84 | 336 | temp[static_cast<size_t>(i)] += a.values[a_idx] * b_val; | |
| 85 | } | ||
| 86 | } | ||
| 87 | |||
| 88 | 224 | return temp; | |
| 89 | } | ||
| 90 | |||
| 91 | int CountNonZero(const std::vector<double> &column, double epsilon) { | ||
| 92 | int count = 0; | ||
| 93 |
4/4✓ Branch 0 taken 168 times.
✓ Branch 1 taken 168 times.
✓ Branch 2 taken 336 times.
✓ Branch 3 taken 112 times.
|
448 | for (double val : column) { |
| 94 |
2/2✓ Branch 0 taken 168 times.
✓ Branch 1 taken 168 times.
|
336 | if (std::abs(val) > epsilon) { |
| 95 | 168 | ++count; | |
| 96 | } | ||
| 97 | } | ||
| 98 | return count; | ||
| 99 | } | ||
| 100 | |||
| 101 | void FillColumn(const std::vector<double> &column, double epsilon, std::vector<int> &row_indices, | ||
| 102 | std::vector<double> &values, int start_pos) { | ||
| 103 | int pos = start_pos; | ||
| 104 |
2/2✓ Branch 0 taken 336 times.
✓ Branch 1 taken 112 times.
|
448 | for (size_t i = 0; i < column.size(); ++i) { |
| 105 |
2/2✓ Branch 0 taken 168 times.
✓ Branch 1 taken 168 times.
|
336 | if (std::abs(column[i]) > epsilon) { |
| 106 | 168 | row_indices[static_cast<size_t>(pos)] = static_cast<int>(i); | |
| 107 | 168 | values[static_cast<size_t>(pos)] = column[i]; | |
| 108 | 168 | ++pos; | |
| 109 | } | ||
| 110 | } | ||
| 111 | } | ||
| 112 | |||
| 113 | } // namespace | ||
| 114 | |||
| 115 | 40 | SparseMatrixCCS KotelnikovaATaskSTL::MultiplyMatrices(const SparseMatrixCCS &a, const SparseMatrixCCS &b) { | |
| 116 | 40 | SparseMatrixCCS result(a.rows, b.cols); | |
| 117 | |||
| 118 | 40 | const double epsilon = 1e-10; | |
| 119 | 40 | const unsigned int hardware_threads = std::thread::hardware_concurrency(); | |
| 120 | 40 | const int num_threads = (hardware_threads > 0) ? static_cast<int>(hardware_threads) : 1; | |
| 121 | 40 | const int cols_per_thread = (b.cols + num_threads - 1) / num_threads; | |
| 122 | |||
| 123 |
1/2✓ Branch 1 taken 40 times.
✗ Branch 2 not taken.
|
40 | std::vector<int> col_start(b.cols, 0); |
| 124 | 40 | std::vector<std::thread> threads; | |
| 125 | |||
| 126 |
2/2✓ Branch 0 taken 160 times.
✓ Branch 1 taken 40 times.
|
200 | for (int thread_idx = 0; thread_idx < num_threads; ++thread_idx) { |
| 127 | 160 | int start_col = thread_idx * cols_per_thread; | |
| 128 |
1/2✓ Branch 1 taken 160 times.
✗ Branch 2 not taken.
|
160 | int end_col = std::min(start_col + cols_per_thread, b.cols); |
| 129 | |||
| 130 |
1/2✓ Branch 1 taken 160 times.
✗ Branch 2 not taken.
|
160 | threads.emplace_back([&, start_col, end_col]() { |
| 131 |
2/2✓ Branch 0 taken 112 times.
✓ Branch 1 taken 160 times.
|
272 | for (int j = start_col; j < end_col; ++j) { |
| 132 | 112 | std::vector<double> column = ComputeColumn(a, b, j); | |
| 133 |
1/2✓ Branch 0 taken 112 times.
✗ Branch 1 not taken.
|
224 | col_start[j] = CountNonZero(column, epsilon); |
| 134 | } | ||
| 135 | 160 | }); | |
| 136 | } | ||
| 137 | |||
| 138 |
2/2✓ Branch 0 taken 160 times.
✓ Branch 1 taken 40 times.
|
200 | for (auto &thread : threads) { |
| 139 |
1/2✓ Branch 1 taken 160 times.
✗ Branch 2 not taken.
|
160 | thread.join(); |
| 140 | } | ||
| 141 | |||
| 142 |
1/2✓ Branch 1 taken 40 times.
✗ Branch 2 not taken.
|
40 | std::vector<int> col_ptr(b.cols + 1, 0); |
| 143 |
2/2✓ Branch 0 taken 112 times.
✓ Branch 1 taken 40 times.
|
152 | for (int j = 0; j < b.cols; ++j) { |
| 144 | 112 | col_ptr[j + 1] = col_ptr[j] + col_start[j]; | |
| 145 | } | ||
| 146 | |||
| 147 |
1/2✓ Branch 1 taken 40 times.
✗ Branch 2 not taken.
|
40 | const int total_nnz = col_ptr[b.cols]; |
| 148 |
1/2✓ Branch 1 taken 40 times.
✗ Branch 2 not taken.
|
40 | result.values.resize(static_cast<size_t>(total_nnz)); |
| 149 |
1/2✓ Branch 1 taken 40 times.
✗ Branch 2 not taken.
|
40 | result.row_indices.resize(static_cast<size_t>(total_nnz)); |
| 150 |
1/2✓ Branch 1 taken 40 times.
✗ Branch 2 not taken.
|
40 | result.col_ptrs = col_ptr; |
| 151 | |||
| 152 | 40 | threads.clear(); | |
| 153 |
2/2✓ Branch 0 taken 160 times.
✓ Branch 1 taken 40 times.
|
200 | for (int thread_idx = 0; thread_idx < num_threads; ++thread_idx) { |
| 154 | 160 | int start_col = thread_idx * cols_per_thread; | |
| 155 |
1/2✓ Branch 1 taken 160 times.
✗ Branch 2 not taken.
|
160 | int end_col = std::min(start_col + cols_per_thread, b.cols); |
| 156 | |||
| 157 |
1/4✓ Branch 1 taken 160 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
|
160 | threads.emplace_back([&, start_col, end_col]() { |
| 158 |
2/2✓ Branch 0 taken 112 times.
✓ Branch 1 taken 160 times.
|
272 | for (int j = start_col; j < end_col; ++j) { |
| 159 | 112 | std::vector<double> column = ComputeColumn(a, b, j); | |
| 160 | 112 | FillColumn(column, epsilon, result.row_indices, result.values, col_ptr[j]); | |
| 161 | } | ||
| 162 | 160 | }); | |
| 163 | } | ||
| 164 | |||
| 165 |
2/2✓ Branch 0 taken 160 times.
✓ Branch 1 taken 40 times.
|
200 | for (auto &thread : threads) { |
| 166 |
1/2✓ Branch 1 taken 160 times.
✗ Branch 2 not taken.
|
160 | thread.join(); |
| 167 | } | ||
| 168 | |||
| 169 | 40 | return result; | |
| 170 | 40 | } | |
| 171 | |||
| 172 | 40 | bool KotelnikovaATaskSTL::RunImpl() { | |
| 173 | const auto &[a, b] = GetInput(); | ||
| 174 | 40 | GetOutput() = MultiplyMatrices(a, b); | |
| 175 | 40 | return true; | |
| 176 | } | ||
| 177 | |||
| 178 | 40 | bool KotelnikovaATaskSTL::PostProcessingImpl() { | |
| 179 | 40 | return true; | |
| 180 | } | ||
| 181 | |||
| 182 | } // namespace kotelnikova_a_double_matr_mult | ||
| 183 |