| Line | Branch | Exec | Source |
|---|---|---|---|
| 1 | #include "olesnitskiy_v_striped_matrix_multiplication/seq/include/ops_seq.hpp" | ||
| 2 | |||
| 3 | #include <algorithm> | ||
| 4 | #include <cstddef> | ||
| 5 | #include <tuple> | ||
| 6 | #include <vector> | ||
| 7 | |||
| 8 | #include "olesnitskiy_v_striped_matrix_multiplication/common/include/common.hpp" | ||
| 9 | |||
| 10 | namespace olesnitskiy_v_striped_matrix_multiplication { | ||
| 11 | |||
| 12 | namespace { | ||
| 13 | 120 | int FindCommonDivisor(int a, int b, int max_divisor) { | |
| 14 |
2/4✓ Branch 0 taken 120 times.
✗ Branch 1 not taken.
✗ Branch 2 not taken.
✓ Branch 3 taken 120 times.
|
120 | if (a <= 0 || b <= 0 || max_divisor <= 1) { |
| 15 | return 1; | ||
| 16 | } | ||
| 17 | |||
| 18 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 224 times.
|
344 | for (int divisor = std::min({a, b, max_divisor}); divisor >= 1; --divisor) { |
| 19 |
4/4✓ Branch 0 taken 128 times.
✓ Branch 1 taken 96 times.
✓ Branch 2 taken 8 times.
✓ Branch 3 taken 120 times.
|
224 | if (a % divisor == 0 && b % divisor == 0) { |
| 20 | return divisor; | ||
| 21 | } | ||
| 22 | } | ||
| 23 | |||
| 24 | return 1; | ||
| 25 | } | ||
| 26 | } // namespace | ||
| 27 | |||
| 28 |
1/2✓ Branch 1 taken 120 times.
✗ Branch 2 not taken.
|
120 | OlesnitskiyVStripedMatrixMultiplicationSEQ::OlesnitskiyVStripedMatrixMultiplicationSEQ(const InType &in) { |
| 29 | SetTypeOfTask(GetStaticTypeOfTask()); | ||
| 30 | GetInput() = in; | ||
| 31 | 120 | GetOutput() = std::make_tuple(0, 0, std::vector<double>()); | |
| 32 | 120 | } | |
| 33 | |||
| 34 | 120 | bool OlesnitskiyVStripedMatrixMultiplicationSEQ::ValidationImpl() { | |
| 35 | const auto &[rows_a, cols_a, data_a, rows_b, cols_b, data_b] = GetInput(); | ||
| 36 | const auto &[out_rows, out_cols, out_data] = GetOutput(); | ||
| 37 | 120 | rows_a_ = rows_a; | |
| 38 | 120 | cols_a_ = cols_a; | |
| 39 | 120 | data_a_ = data_a; | |
| 40 | 120 | rows_b_ = rows_b; | |
| 41 | 120 | cols_b_ = cols_b; | |
| 42 | 120 | data_b_ = data_b; | |
| 43 |
4/8✓ Branch 0 taken 120 times.
✗ Branch 1 not taken.
✓ Branch 2 taken 120 times.
✗ Branch 3 not taken.
✓ Branch 4 taken 120 times.
✗ Branch 5 not taken.
✓ Branch 6 taken 120 times.
✗ Branch 7 not taken.
|
120 | if (rows_a == 0 || cols_a == 0 || rows_b == 0 || cols_b == 0) { |
| 44 | return false; | ||
| 45 | } | ||
| 46 |
2/4✓ Branch 0 taken 120 times.
✗ Branch 1 not taken.
✓ Branch 2 taken 120 times.
✗ Branch 3 not taken.
|
120 | if (data_a.size() != rows_a * cols_a || data_b.size() != rows_b * cols_b) { |
| 47 | return false; | ||
| 48 | } | ||
| 49 |
1/2✓ Branch 0 taken 120 times.
✗ Branch 1 not taken.
|
120 | if (cols_a != rows_b) { |
| 50 | return false; | ||
| 51 | } | ||
| 52 |
3/6✓ Branch 0 taken 120 times.
✗ Branch 1 not taken.
✓ Branch 2 taken 120 times.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
✓ Branch 5 taken 120 times.
|
120 | if (out_rows != 0 || out_cols != 0 || !out_data.empty()) { |
| 53 | ✗ | return false; | |
| 54 | } | ||
| 55 | |||
| 56 | return true; | ||
| 57 | } | ||
| 58 | |||
| 59 | 120 | bool OlesnitskiyVStripedMatrixMultiplicationSEQ::PreProcessingImpl() { | |
| 60 | 120 | rows_c_ = rows_a_; | |
| 61 | 120 | cols_c_ = cols_b_; | |
| 62 | int max_stripes = 8; | ||
| 63 | 120 | num_stripes_ = FindCommonDivisor(static_cast<int>(rows_a_), static_cast<int>(cols_b_), max_stripes); | |
| 64 |
2/2✓ Branch 0 taken 24 times.
✓ Branch 1 taken 96 times.
|
120 | if (num_stripes_ < 2) { |
| 65 | 24 | num_stripes_ = 1; | |
| 66 | } | ||
| 67 | 120 | result_c_.resize(rows_c_ * cols_c_, 0.0); | |
| 68 | 120 | GetOutput() = std::make_tuple(0, 0, std::vector<double>()); | |
| 69 | 120 | return true; | |
| 70 | } | ||
| 71 | |||
| 72 | 120 | bool OlesnitskiyVStripedMatrixMultiplicationSEQ::RunImpl() { | |
| 73 |
2/4✓ Branch 0 taken 120 times.
✗ Branch 1 not taken.
✗ Branch 2 not taken.
✓ Branch 3 taken 120 times.
|
120 | if (rows_a_ == 0 || cols_b_ == 0) { |
| 74 | ✗ | GetOutput() = std::make_tuple(rows_c_, cols_c_, std::vector<double>()); | |
| 75 | ✗ | return true; | |
| 76 | } | ||
| 77 | |||
| 78 | 120 | result_c_.resize(rows_c_ * cols_c_, 0.0); | |
| 79 | |||
| 80 | bool success = false; | ||
| 81 |
2/2✓ Branch 0 taken 24 times.
✓ Branch 1 taken 96 times.
|
120 | if (num_stripes_ == 1) { |
| 82 | 24 | success = MultiplySimple(); | |
| 83 | } else { | ||
| 84 | 96 | success = MultiplyStriped(); | |
| 85 | } | ||
| 86 | |||
| 87 |
1/2✓ Branch 0 taken 96 times.
✗ Branch 1 not taken.
|
96 | if (success) { |
| 88 | 120 | GetOutput() = std::make_tuple(rows_c_, cols_c_, result_c_); | |
| 89 | } | ||
| 90 | |||
| 91 | return success; | ||
| 92 | } | ||
| 93 | |||
| 94 | 24 | bool OlesnitskiyVStripedMatrixMultiplicationSEQ::MultiplySimple() { | |
| 95 |
2/2✓ Branch 0 taken 64 times.
✓ Branch 1 taken 24 times.
|
88 | for (size_t i = 0; i < rows_a_; ++i) { |
| 96 |
2/2✓ Branch 0 taken 152 times.
✓ Branch 1 taken 64 times.
|
216 | for (size_t j = 0; j < cols_b_; ++j) { |
| 97 | double sum = 0.0; | ||
| 98 |
2/2✓ Branch 0 taken 776 times.
✓ Branch 1 taken 152 times.
|
928 | for (size_t k = 0; k < cols_a_; ++k) { |
| 99 | 776 | sum += data_a_[(i * cols_a_) + k] * data_b_[(k * cols_b_) + j]; | |
| 100 | } | ||
| 101 | 152 | result_c_[(i * cols_b_) + j] = sum; | |
| 102 | } | ||
| 103 | } | ||
| 104 | 24 | return true; | |
| 105 | } | ||
| 106 | |||
| 107 | 2288 | bool OlesnitskiyVStripedMatrixMultiplicationSEQ::ProcessStripePair(int stripe_a, int stripe_b, size_t rows_per_stripe, | |
| 108 | size_t cols_per_stripe) { | ||
| 109 | 2288 | const size_t start_row_a = static_cast<size_t>(stripe_a) * rows_per_stripe; | |
| 110 | 2288 | const size_t start_col_b = static_cast<size_t>(stripe_b) * cols_per_stripe; | |
| 111 | |||
| 112 |
2/2✓ Branch 0 taken 2920 times.
✓ Branch 1 taken 2288 times.
|
5208 | for (size_t i = 0; i < rows_per_stripe; ++i) { |
| 113 | 2920 | const size_t row_idx = start_row_a + i; | |
| 114 |
2/2✓ Branch 0 taken 4328 times.
✓ Branch 1 taken 2920 times.
|
7248 | for (size_t j = 0; j < cols_per_stripe; ++j) { |
| 115 | 4328 | const size_t col_idx = start_col_b + j; | |
| 116 | double sum = 0.0; | ||
| 117 | |||
| 118 |
2/2✓ Branch 0 taken 38328 times.
✓ Branch 1 taken 4328 times.
|
42656 | for (size_t k = 0; k < cols_a_; ++k) { |
| 119 | 38328 | sum += data_a_[(row_idx * cols_a_) + k] * data_b_[(k * cols_b_) + col_idx]; | |
| 120 | } | ||
| 121 | 4328 | result_c_[(row_idx * cols_b_) + col_idx] = sum; | |
| 122 | } | ||
| 123 | } | ||
| 124 | 2288 | return true; | |
| 125 | } | ||
| 126 | |||
| 127 | 96 | bool OlesnitskiyVStripedMatrixMultiplicationSEQ::MultiplyStriped() { | |
| 128 |
2/4✓ Branch 0 taken 96 times.
✗ Branch 1 not taken.
✓ Branch 2 taken 96 times.
✗ Branch 3 not taken.
|
96 | if (rows_a_ % static_cast<size_t>(num_stripes_) != 0 || cols_b_ % static_cast<size_t>(num_stripes_) != 0) { |
| 129 | return false; | ||
| 130 | } | ||
| 131 | |||
| 132 | 96 | const size_t rows_per_stripe = rows_a_ / static_cast<size_t>(num_stripes_); | |
| 133 | 96 | const size_t cols_per_stripe = cols_b_ / static_cast<size_t>(num_stripes_); | |
| 134 | |||
| 135 |
2/2✓ Branch 0 taken 432 times.
✓ Branch 1 taken 96 times.
|
528 | for (int stripe_a = 0; stripe_a < num_stripes_; ++stripe_a) { |
| 136 |
2/2✓ Branch 0 taken 2288 times.
✓ Branch 1 taken 432 times.
|
2720 | for (int stripe_b = 0; stripe_b < num_stripes_; ++stripe_b) { |
| 137 | 2288 | if (!ProcessStripePair(stripe_a, stripe_b, rows_per_stripe, cols_per_stripe)) { | |
| 138 | return false; | ||
| 139 | } | ||
| 140 | } | ||
| 141 | } | ||
| 142 | return true; | ||
| 143 | } | ||
| 144 | |||
| 145 | 120 | bool OlesnitskiyVStripedMatrixMultiplicationSEQ::PostProcessingImpl() { | |
| 146 | 120 | return true; | |
| 147 | } | ||
| 148 | } // namespace olesnitskiy_v_striped_matrix_multiplication | ||
| 149 |