| Line | Branch | Exec | Source |
|---|---|---|---|
| 1 | #include "safronov_m_multiplication_matrix_blockscheme_cannon/omp/include/ops_omp.hpp" | ||
| 2 | |||
| 3 | #include <omp.h> | ||
| 4 | |||
| 5 | #include <utility> | ||
| 6 | #include <vector> | ||
| 7 | |||
| 8 | #include "safronov_m_multiplication_matrix_blockscheme_cannon/common/include/common.hpp" | ||
| 9 | |||
| 10 | namespace safronov_m_multiplication_matrix_blocksscheme_cannon { | ||
| 11 | |||
| 12 |
1/2✓ Branch 1 taken 32 times.
✗ Branch 2 not taken.
|
32 | SafronovMMultiplicationMatrixBlockSchemeCannonOMP::SafronovMMultiplicationMatrixBlockSchemeCannonOMP(const InType &in) { |
| 13 | SetTypeOfTask(GetStaticTypeOfTask()); | ||
| 14 | GetInput() = in; | ||
| 15 | 32 | } | |
| 16 | |||
| 17 | 32 | bool SafronovMMultiplicationMatrixBlockSchemeCannonOMP::ValidationImpl() { | |
| 18 | const auto &in = GetInput(); | ||
| 19 | 32 | int size_block = std::get<0>(in); | |
| 20 | const auto &matrix_a = std::get<1>(in); | ||
| 21 | const auto &matrix_b = std::get<2>(in); | ||
| 22 |
4/8✗ Branch 0 not taken.
✓ Branch 1 taken 32 times.
✗ Branch 2 not taken.
✓ Branch 3 taken 32 times.
✗ Branch 4 not taken.
✓ Branch 5 taken 32 times.
✗ Branch 6 not taken.
✓ Branch 7 taken 32 times.
|
32 | return (size_block > 0) && (!matrix_a.empty() && !matrix_b.empty()) && (matrix_a.size() == matrix_a[0].size()) && |
| 23 |
2/4✓ Branch 0 taken 32 times.
✗ Branch 1 not taken.
✗ Branch 2 not taken.
✓ Branch 3 taken 32 times.
|
64 | (matrix_b.size() == matrix_b[0].size()) && (matrix_a.size() == matrix_b.size()) && |
| 24 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 32 times.
|
32 | (matrix_a.size() % size_block == 0); |
| 25 | } | ||
| 26 | |||
| 27 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 32 times.
|
32 | bool SafronovMMultiplicationMatrixBlockSchemeCannonOMP::PreProcessingImpl() { |
| 28 | GetOutput().clear(); | ||
| 29 | 32 | return true; | |
| 30 | } | ||
| 31 | |||
| 32 | 304 | void SafronovMMultiplicationMatrixBlockSchemeCannonOMP::MultiplyingBlocks(std::vector<std::vector<double>> &block_a, | |
| 33 | std::vector<std::vector<double>> &block_b, | ||
| 34 | std::vector<std::vector<double>> &block_c, | ||
| 35 | int size_block) { | ||
| 36 |
2/2✓ Branch 0 taken 784 times.
✓ Branch 1 taken 304 times.
|
1088 | for (int i = 0; i < size_block; i++) { |
| 37 |
2/2✓ Branch 0 taken 2224 times.
✓ Branch 1 taken 784 times.
|
3008 | for (int j = 0; j < size_block; j++) { |
| 38 |
2/2✓ Branch 0 taken 6736 times.
✓ Branch 1 taken 2224 times.
|
8960 | for (int k = 0; k < size_block; k++) { |
| 39 | 6736 | block_c[i][j] += block_a[i][k] * block_b[k][j]; | |
| 40 | } | ||
| 41 | } | ||
| 42 | } | ||
| 43 | 304 | } | |
| 44 | |||
| 45 | ✗ | void SafronovMMultiplicationMatrixBlockSchemeCannonOMP::ShiftingBlocksMatrixALeft( | |
| 46 | std::vector<std::vector<std::vector<std::vector<double>>>> &matrix_blocks_a, int columns) { | ||
| 47 | 32 | #pragma omp parallel for default(none) shared(matrix_blocks_a, columns) | |
| 48 | for (int i = 0; i < columns; i++) { | ||
| 49 | std::vector<std::vector<double>> tmp = std::move(matrix_blocks_a[i][0]); | ||
| 50 | for (int j = 1; j < columns; j++) { | ||
| 51 | matrix_blocks_a[i][j - 1] = std::move(matrix_blocks_a[i][j]); | ||
| 52 | } | ||
| 53 | matrix_blocks_a[i][columns - 1] = std::move(tmp); | ||
| 54 | } | ||
| 55 | ✗ | } | |
| 56 | |||
| 57 | ✗ | void SafronovMMultiplicationMatrixBlockSchemeCannonOMP::ShiftingBlocksMatrixBUp( | |
| 58 | std::vector<std::vector<std::vector<std::vector<double>>>> &matrix_blocks_b, int columns) { | ||
| 59 | 32 | #pragma omp parallel for default(none) shared(matrix_blocks_b, columns) | |
| 60 | for (int i = 0; i < columns; i++) { | ||
| 61 | std::vector<std::vector<double>> tmp = std::move(matrix_blocks_b[0][i]); | ||
| 62 | for (int j = 1; j < columns; j++) { | ||
| 63 | matrix_blocks_b[j - 1][i] = std::move(matrix_blocks_b[j][i]); | ||
| 64 | } | ||
| 65 | matrix_blocks_b[columns - 1][i] = std::move(tmp); | ||
| 66 | } | ||
| 67 | 32 | } | |
| 68 | |||
| 69 | 32 | void SafronovMMultiplicationMatrixBlockSchemeCannonOMP::AlgorithmCannon( | |
| 70 | std::vector<std::vector<std::vector<std::vector<double>>>> &matrix_blocks_a, | ||
| 71 | std::vector<std::vector<std::vector<std::vector<double>>>> &matrix_blocks_b, | ||
| 72 | std::vector<std::vector<std::vector<std::vector<double>>>> &matrix_blocks_c, int size_block, int columns_blocks) { | ||
| 73 |
2/2✓ Branch 0 taken 64 times.
✓ Branch 1 taken 32 times.
|
96 | for (int i = 0; i < columns_blocks; i++) { |
| 74 | 64 | #pragma omp parallel for collapse(2) default(none) \ | |
| 75 | shared(matrix_blocks_a, matrix_blocks_b, matrix_blocks_c, size_block, columns_blocks) | ||
| 76 | for (int j = 0; j < columns_blocks; j++) { | ||
| 77 | for (int k = 0; k < columns_blocks; k++) { | ||
| 78 | SafronovMMultiplicationMatrixBlockSchemeCannonOMP::MultiplyingBlocks( | ||
| 79 | matrix_blocks_a[j][k], matrix_blocks_b[j][k], matrix_blocks_c[j][k], size_block); | ||
| 80 | } | ||
| 81 | } | ||
| 82 | |||
| 83 |
2/2✓ Branch 0 taken 32 times.
✓ Branch 1 taken 32 times.
|
64 | if (i < columns_blocks - 1) { |
| 84 | SafronovMMultiplicationMatrixBlockSchemeCannonOMP::ShiftingBlocksMatrixALeft(matrix_blocks_a, columns_blocks); | ||
| 85 | SafronovMMultiplicationMatrixBlockSchemeCannonOMP::ShiftingBlocksMatrixBUp(matrix_blocks_b, columns_blocks); | ||
| 86 | } | ||
| 87 | } | ||
| 88 | 32 | } | |
| 89 | |||
| 90 | ✗ | void SafronovMMultiplicationMatrixBlockSchemeCannonOMP::FillingResultingMatrix( | |
| 91 | std::vector<std::vector<std::vector<std::vector<double>>>> &matrix_blocks_c, | ||
| 92 | std::vector<std::vector<double>> &matrix_c, int size_block, int columns_blocks) { | ||
| 93 | 32 | #pragma omp parallel for collapse(2) default(none) shared(matrix_blocks_c, matrix_c, size_block, columns_blocks) | |
| 94 | for (int i = 0; i < columns_blocks; i++) { | ||
| 95 | for (int j = 0; j < columns_blocks; j++) { | ||
| 96 | for (int k = 0; k < size_block; k++) { | ||
| 97 | for (int col = 0; col < size_block; col++) { | ||
| 98 | matrix_c[(i * size_block) + k][(j * size_block) + col] = matrix_blocks_c[i][j][k][col]; | ||
| 99 | } | ||
| 100 | } | ||
| 101 | } | ||
| 102 | } | ||
| 103 | ✗ | } | |
| 104 | |||
| 105 | 32 | bool SafronovMMultiplicationMatrixBlockSchemeCannonOMP::RunImpl() { | |
| 106 | const auto &in = GetInput(); | ||
| 107 | 32 | int size_block = std::get<0>(in); | |
| 108 | const auto &matrix_a = std::get<1>(in); | ||
| 109 | const auto &matrix_b = std::get<2>(in); | ||
| 110 | 32 | int n = static_cast<int>(matrix_a.size()); | |
| 111 | 32 | int columns_blocks = n / size_block; | |
| 112 | std::vector<std::vector<std::vector<std::vector<double>>>> matrix_blocks_a( | ||
| 113 | columns_blocks, | ||
| 114 | 32 | std::vector<std::vector<std::vector<double>>>( | |
| 115 |
3/6✓ Branch 2 taken 32 times.
✗ Branch 3 not taken.
✓ Branch 5 taken 32 times.
✗ Branch 6 not taken.
✓ Branch 8 taken 32 times.
✗ Branch 9 not taken.
|
64 | columns_blocks, std::vector<std::vector<double>>(size_block, std::vector<double>(size_block)))); |
| 116 | std::vector<std::vector<std::vector<std::vector<double>>>> matrix_blocks_b( | ||
| 117 | columns_blocks, | ||
| 118 | 32 | std::vector<std::vector<std::vector<double>>>( | |
| 119 |
4/8✓ Branch 1 taken 32 times.
✗ Branch 2 not taken.
✓ Branch 4 taken 32 times.
✗ Branch 5 not taken.
✓ Branch 7 taken 32 times.
✗ Branch 8 not taken.
✓ Branch 10 taken 32 times.
✗ Branch 11 not taken.
|
64 | columns_blocks, std::vector<std::vector<double>>(size_block, std::vector<double>(size_block)))); |
| 120 | std::vector<std::vector<std::vector<std::vector<double>>>> matrix_blocks_c( | ||
| 121 | columns_blocks, | ||
| 122 | 32 | std::vector<std::vector<std::vector<double>>>( | |
| 123 |
4/8✓ Branch 1 taken 32 times.
✗ Branch 2 not taken.
✓ Branch 4 taken 32 times.
✗ Branch 5 not taken.
✓ Branch 7 taken 32 times.
✗ Branch 8 not taken.
✓ Branch 10 taken 32 times.
✗ Branch 11 not taken.
|
64 | columns_blocks, std::vector<std::vector<double>>(size_block, std::vector<double>(size_block, 0.0)))); |
| 124 | |||
| 125 | 32 | #pragma omp parallel for collapse(2) default(none) \ | |
| 126 | shared(matrix_blocks_a, matrix_blocks_b, matrix_a, matrix_b, size_block, columns_blocks) | ||
| 127 | for (int i = 0; i < columns_blocks; i++) { | ||
| 128 | for (int j = 0; j < columns_blocks; j++) { | ||
| 129 | int shift = (i + j) % columns_blocks; | ||
| 130 | for (int k = 0; k < size_block; k++) { | ||
| 131 | for (int col = 0; col < size_block; col++) { | ||
| 132 | matrix_blocks_a[i][j][k][col] = matrix_a[(i * size_block) + k][(shift * size_block) + col]; | ||
| 133 | matrix_blocks_b[i][j][k][col] = matrix_b[(shift * size_block) + k][(j * size_block) + col]; | ||
| 134 | } | ||
| 135 | } | ||
| 136 | } | ||
| 137 | } | ||
| 138 | 32 | SafronovMMultiplicationMatrixBlockSchemeCannonOMP::AlgorithmCannon(matrix_blocks_a, matrix_blocks_b, matrix_blocks_c, | |
| 139 | size_block, columns_blocks); | ||
| 140 | |||
| 141 |
2/4✓ Branch 1 taken 32 times.
✗ Branch 2 not taken.
✓ Branch 4 taken 32 times.
✗ Branch 5 not taken.
|
64 | std::vector<std::vector<double>> matrix_c(n, std::vector<double>(n)); |
| 142 | SafronovMMultiplicationMatrixBlockSchemeCannonOMP::FillingResultingMatrix(matrix_blocks_c, matrix_c, size_block, | ||
| 143 | columns_blocks); | ||
| 144 | 32 | GetOutput() = std::move(matrix_c); | |
| 145 | 32 | return true; | |
| 146 | 32 | } | |
| 147 | |||
| 148 | 32 | bool SafronovMMultiplicationMatrixBlockSchemeCannonOMP::PostProcessingImpl() { | |
| 149 | 32 | return true; | |
| 150 | } | ||
| 151 | |||
| 152 | } // namespace safronov_m_multiplication_matrix_blocksscheme_cannon | ||
| 153 |