| Line | Branch | Exec | Source |
|---|---|---|---|
| 1 | #include "viderman_a_sparse_matrix_mult_crs_complex/tbb/include/ops_tbb.hpp" | ||
| 2 | |||
| 3 | #include <algorithm> | ||
| 4 | #include <cmath> | ||
| 5 | #include <cstddef> | ||
| 6 | #include <vector> | ||
| 7 | |||
| 8 | #include "oneapi/tbb/enumerable_thread_specific.h" | ||
| 9 | #include "oneapi/tbb/parallel_for.h" | ||
| 10 | #include "viderman_a_sparse_matrix_mult_crs_complex/common/include/common.hpp" | ||
| 11 | |||
| 12 | namespace viderman_a_sparse_matrix_mult_crs_complex { | ||
| 13 | namespace { | ||
| 14 | |||
| 15 |
2/2✓ Branch 0 taken 4 times.
✓ Branch 1 taken 36 times.
|
40 | void ProcessRow(const CRSMatrix &a, const CRSMatrix &b, int row, std::vector<Complex> &accumulator, |
| 16 | std::vector<int> &marker, std::vector<int> ¤t_row_indices, std::vector<int> &dst_cols, | ||
| 17 | std::vector<Complex> &dst_vals) { | ||
| 18 | current_row_indices.clear(); | ||
| 19 | |||
| 20 |
2/2✓ Branch 0 taken 28 times.
✓ Branch 1 taken 40 times.
|
68 | for (int j = a.row_ptr[row]; j < a.row_ptr[row + 1]; ++j) { |
| 21 | 28 | const int col_a = a.col_indices[j]; | |
| 22 | 28 | const Complex val_a = a.values[j]; | |
| 23 | |||
| 24 |
2/2✓ Branch 0 taken 28 times.
✓ Branch 1 taken 28 times.
|
56 | for (int k = b.row_ptr[col_a]; k < b.row_ptr[col_a + 1]; ++k) { |
| 25 |
1/2✓ Branch 0 taken 28 times.
✗ Branch 1 not taken.
|
28 | const int col_b = b.col_indices[k]; |
| 26 |
1/2✓ Branch 0 taken 28 times.
✗ Branch 1 not taken.
|
28 | accumulator[col_b] += val_a * b.values[k]; |
| 27 | |||
| 28 |
1/2✓ Branch 0 taken 28 times.
✗ Branch 1 not taken.
|
28 | if (marker[col_b] != row) { |
| 29 | current_row_indices.push_back(col_b); | ||
| 30 | 28 | marker[col_b] = row; | |
| 31 | } | ||
| 32 | } | ||
| 33 | } | ||
| 34 | |||
| 35 | std::ranges::sort(current_row_indices); | ||
| 36 | |||
| 37 | dst_cols.clear(); | ||
| 38 | dst_vals.clear(); | ||
| 39 | 40 | dst_cols.reserve(current_row_indices.size()); | |
| 40 | 40 | dst_vals.reserve(current_row_indices.size()); | |
| 41 | |||
| 42 |
2/2✓ Branch 0 taken 28 times.
✓ Branch 1 taken 40 times.
|
68 | for (const int idx : current_row_indices) { |
| 43 |
1/2✓ Branch 0 taken 28 times.
✗ Branch 1 not taken.
|
28 | if (std::abs(accumulator[idx]) > kEpsilon) { |
| 44 | dst_cols.push_back(idx); | ||
| 45 | dst_vals.push_back(accumulator[idx]); | ||
| 46 | } | ||
| 47 | 28 | accumulator[idx] = Complex(0.0, 0.0); | |
| 48 | } | ||
| 49 | 40 | } | |
| 50 | |||
| 51 | struct ThreadBuffers { | ||
| 52 | std::vector<Complex> accumulator; | ||
| 53 | std::vector<int> marker; | ||
| 54 | std::vector<int> current_row_indices; | ||
| 55 | }; | ||
| 56 | |||
| 57 | } // namespace | ||
| 58 | |||
| 59 |
1/2✓ Branch 2 taken 24 times.
✗ Branch 3 not taken.
|
24 | VidermanASparseMatrixMultCRSComplexTBB::VidermanASparseMatrixMultCRSComplexTBB(const InType &in) { |
| 60 | SetTypeOfTask(GetStaticTypeOfTask()); | ||
| 61 | GetInput() = in; | ||
| 62 | 24 | GetOutput() = CRSMatrix(); | |
| 63 | 24 | } | |
| 64 | |||
| 65 |
1/2✓ Branch 0 taken 24 times.
✗ Branch 1 not taken.
|
24 | bool VidermanASparseMatrixMultCRSComplexTBB::ValidationImpl() { |
| 66 | const auto &input = GetInput(); | ||
| 67 | const auto &a = std::get<0>(input); | ||
| 68 | const auto &b = std::get<1>(input); | ||
| 69 | |||
| 70 |
2/4✓ Branch 0 taken 24 times.
✗ Branch 1 not taken.
✓ Branch 2 taken 24 times.
✗ Branch 3 not taken.
|
24 | if (!a.IsValid() || !b.IsValid()) { |
| 71 | return false; | ||
| 72 | } | ||
| 73 | |||
| 74 |
2/2✓ Branch 0 taken 4 times.
✓ Branch 1 taken 20 times.
|
24 | if (a.cols != b.rows) { |
| 75 | 4 | return false; | |
| 76 | } | ||
| 77 | |||
| 78 | return true; | ||
| 79 | } | ||
| 80 | |||
| 81 | 20 | bool VidermanASparseMatrixMultCRSComplexTBB::PreProcessingImpl() { | |
| 82 | const auto &input = GetInput(); | ||
| 83 | |||
| 84 | 20 | a_ = &std::get<0>(input); | |
| 85 | 20 | b_ = &std::get<1>(input); | |
| 86 | |||
| 87 | 20 | return true; | |
| 88 | } | ||
| 89 | |||
| 90 | 20 | void VidermanASparseMatrixMultCRSComplexTBB::Multiply(const CRSMatrix &a, const CRSMatrix &b, CRSMatrix &c) { | |
| 91 | 20 | c.rows = a.rows; | |
| 92 | 20 | c.cols = b.cols; | |
| 93 |
1/2✗ Branch 1 not taken.
✓ Branch 2 taken 20 times.
|
20 | c.row_ptr.assign(a.rows + 1, 0); |
| 94 | c.col_indices.clear(); | ||
| 95 | c.values.clear(); | ||
| 96 | |||
| 97 | 20 | std::vector<std::vector<int>> row_cols(a.rows); | |
| 98 |
1/2✓ Branch 1 taken 20 times.
✗ Branch 2 not taken.
|
20 | std::vector<std::vector<Complex>> row_vals(a.rows); |
| 99 | |||
| 100 | 24 | oneapi::tbb::enumerable_thread_specific<ThreadBuffers> tls([&b] { | |
| 101 | 24 | ThreadBuffers buffers; | |
| 102 |
1/2✓ Branch 1 taken 24 times.
✗ Branch 2 not taken.
|
24 | buffers.accumulator.assign(b.cols, Complex(0.0, 0.0)); |
| 103 |
1/2✓ Branch 1 taken 24 times.
✗ Branch 2 not taken.
|
24 | buffers.marker.assign(b.cols, -1); |
| 104 | 24 | return buffers; | |
| 105 |
1/2✓ Branch 1 taken 20 times.
✗ Branch 2 not taken.
|
20 | }); |
| 106 | |||
| 107 |
1/2✓ Branch 1 taken 20 times.
✗ Branch 2 not taken.
|
20 | oneapi::tbb::parallel_for(0, a.rows, [&](int i) { |
| 108 | 40 | auto &buffers = tls.local(); | |
| 109 | 40 | ProcessRow(a, b, i, buffers.accumulator, buffers.marker, buffers.current_row_indices, row_cols[i], row_vals[i]); | |
| 110 | 40 | }); | |
| 111 | |||
| 112 |
2/2✓ Branch 0 taken 40 times.
✓ Branch 1 taken 20 times.
|
60 | for (int i = 0; i < a.rows; ++i) { |
| 113 | 40 | c.row_ptr[i + 1] = c.row_ptr[i] + static_cast<int>(row_cols[i].size()); | |
| 114 | } | ||
| 115 | |||
| 116 |
1/2✓ Branch 1 taken 20 times.
✗ Branch 2 not taken.
|
20 | c.col_indices.reserve(static_cast<std::size_t>(c.row_ptr[a.rows])); |
| 117 |
1/2✓ Branch 1 taken 20 times.
✗ Branch 2 not taken.
|
20 | c.values.reserve(static_cast<std::size_t>(c.row_ptr[a.rows])); |
| 118 |
2/2✓ Branch 0 taken 40 times.
✓ Branch 1 taken 20 times.
|
60 | for (int i = 0; i < a.rows; ++i) { |
| 119 |
2/4✓ Branch 1 taken 40 times.
✗ Branch 2 not taken.
✓ Branch 4 taken 40 times.
✗ Branch 5 not taken.
|
40 | c.col_indices.insert(c.col_indices.end(), row_cols[i].begin(), row_cols[i].end()); |
| 120 | 40 | c.values.insert(c.values.end(), row_vals[i].begin(), row_vals[i].end()); | |
| 121 | } | ||
| 122 | 20 | } | |
| 123 | |||
| 124 | 20 | bool VidermanASparseMatrixMultCRSComplexTBB::RunImpl() { | |
| 125 |
2/4✓ Branch 0 taken 20 times.
✗ Branch 1 not taken.
✓ Branch 2 taken 20 times.
✗ Branch 3 not taken.
|
20 | if (a_ == nullptr || b_ == nullptr) { |
| 126 | return false; | ||
| 127 | } | ||
| 128 | |||
| 129 | CRSMatrix &c = GetOutput(); | ||
| 130 | 20 | Multiply(*a_, *b_, c); | |
| 131 | |||
| 132 | 20 | return true; | |
| 133 | } | ||
| 134 | |||
| 135 | 20 | bool VidermanASparseMatrixMultCRSComplexTBB::PostProcessingImpl() { | |
| 136 | CRSMatrix &c = GetOutput(); | ||
| 137 | 20 | return c.IsValid(); | |
| 138 | } | ||
| 139 | |||
| 140 | } // namespace viderman_a_sparse_matrix_mult_crs_complex | ||
| 141 |