| Line | Branch | Exec | Source |
|---|---|---|---|
| 1 | #include "baranov_a_mult_matrix_fox_algorithm/seq/include/ops_seq.hpp" | ||
| 2 | |||
| 3 | #include <algorithm> | ||
| 4 | #include <cmath> | ||
| 5 | #include <cstddef> | ||
| 6 | #include <vector> | ||
| 7 | |||
| 8 | #include "baranov_a_mult_matrix_fox_algorithm/common/include/common.hpp" | ||
| 9 | |||
| 10 | namespace { | ||
| 11 | |||
| 12 | 72 | void ProcessBlock(const std::vector<double> &matrix_a, const std::vector<double> &matrix_b, std::vector<double> &output, | |
| 13 | size_t n, size_t i_start, size_t i_end, size_t j_start, size_t j_end, size_t k_start, size_t k_end) { | ||
| 14 |
2/2✓ Branch 0 taken 4608 times.
✓ Branch 1 taken 72 times.
|
4680 | for (size_t i = i_start; i < i_end; ++i) { |
| 15 |
2/2✓ Branch 0 taken 294912 times.
✓ Branch 1 taken 4608 times.
|
299520 | for (size_t j = j_start; j < j_end; ++j) { |
| 16 | double sum = 0.0; | ||
| 17 |
2/2✓ Branch 0 taken 18874368 times.
✓ Branch 1 taken 294912 times.
|
19169280 | for (size_t k = k_start; k < k_end; ++k) { |
| 18 | 18874368 | sum += matrix_a[(i * n) + k] * matrix_b[(k * n) + j]; | |
| 19 | } | ||
| 20 | 294912 | output[(i * n) + j] += sum; | |
| 21 | } | ||
| 22 | } | ||
| 23 | 72 | } | |
| 24 | |||
| 25 | } // namespace | ||
| 26 | |||
| 27 | namespace baranov_a_mult_matrix_fox_algorithm_seq { | ||
| 28 | |||
| 29 | 160 | BaranovAMultMatrixFoxAlgorithmSEQ::BaranovAMultMatrixFoxAlgorithmSEQ( | |
| 30 |
1/2✓ Branch 1 taken 160 times.
✗ Branch 2 not taken.
|
160 | const baranov_a_mult_matrix_fox_algorithm::InType &in) { |
| 31 | SetTypeOfTask(GetStaticTypeOfTask()); | ||
| 32 | GetInput() = in; | ||
| 33 | 160 | GetOutput() = std::vector<double>(); | |
| 34 | 160 | } | |
| 35 | |||
| 36 | 160 | bool BaranovAMultMatrixFoxAlgorithmSEQ::ValidationImpl() { | |
| 37 | const auto &[matrix_size, matrix_a, matrix_b] = GetInput(); | ||
| 38 | |||
| 39 |
3/6✓ Branch 0 taken 160 times.
✗ Branch 1 not taken.
✗ Branch 2 not taken.
✓ Branch 3 taken 160 times.
✗ Branch 4 not taken.
✓ Branch 5 taken 160 times.
|
160 | return matrix_size > 0 && matrix_a.size() == matrix_size * matrix_size && |
| 40 | 160 | matrix_b.size() == matrix_size * matrix_size; | |
| 41 | } | ||
| 42 | |||
| 43 | 160 | bool BaranovAMultMatrixFoxAlgorithmSEQ::PreProcessingImpl() { | |
| 44 | const auto &[matrix_size, matrix_a, matrix_b] = GetInput(); | ||
| 45 | 160 | GetOutput() = std::vector<double>(matrix_size * matrix_size, 0.0); | |
| 46 | 160 | return true; | |
| 47 | } | ||
| 48 | |||
| 49 | 144 | void BaranovAMultMatrixFoxAlgorithmSEQ::StandardMultiplication(size_t n) { | |
| 50 | const auto &[matrix_size, matrix_a, matrix_b] = GetInput(); | ||
| 51 | auto &output = GetOutput(); | ||
| 52 | |||
| 53 |
2/2✓ Branch 0 taken 960 times.
✓ Branch 1 taken 144 times.
|
1104 | for (size_t i = 0; i < n; ++i) { |
| 54 |
2/2✓ Branch 0 taken 13408 times.
✓ Branch 1 taken 960 times.
|
14368 | for (size_t j = 0; j < n; ++j) { |
| 55 | double sum = 0.0; | ||
| 56 |
2/2✓ Branch 0 taken 314640 times.
✓ Branch 1 taken 13408 times.
|
328048 | for (size_t k = 0; k < n; ++k) { |
| 57 | 314640 | sum += matrix_a[(i * n) + k] * matrix_b[(k * n) + j]; | |
| 58 | } | ||
| 59 | 13408 | output[(i * n) + j] = sum; | |
| 60 | } | ||
| 61 | } | ||
| 62 | 144 | } | |
| 63 | |||
| 64 | 16 | void BaranovAMultMatrixFoxAlgorithmSEQ::FoxBlockMultiplication(size_t n, size_t block_size) { | |
| 65 | const auto &[matrix_size, matrix_a, matrix_b] = GetInput(); | ||
| 66 | auto &output = GetOutput(); | ||
| 67 | |||
| 68 | 16 | size_t num_blocks = (n + block_size - 1) / block_size; | |
| 69 | |||
| 70 |
2/2✓ Branch 0 taken 163840 times.
✓ Branch 1 taken 16 times.
|
163856 | for (size_t idx = 0; idx < n * n; ++idx) { |
| 71 | 163840 | output[idx] = 0.0; | |
| 72 | } | ||
| 73 | |||
| 74 |
2/2✓ Branch 0 taken 24 times.
✓ Branch 1 taken 16 times.
|
40 | for (size_t bi = 0; bi < num_blocks; ++bi) { |
| 75 |
2/2✓ Branch 0 taken 40 times.
✓ Branch 1 taken 24 times.
|
64 | for (size_t bj = 0; bj < num_blocks; ++bj) { |
| 76 |
2/2✓ Branch 0 taken 72 times.
✓ Branch 1 taken 40 times.
|
112 | for (size_t bk = 0; bk < num_blocks; ++bk) { |
| 77 | 72 | size_t broadcast_block = (bi + bk) % num_blocks; | |
| 78 | |||
| 79 | 72 | size_t i_start = bi * block_size; | |
| 80 | 72 | size_t i_end = std::min(i_start + block_size, n); | |
| 81 | 72 | size_t j_start = bj * block_size; | |
| 82 | 72 | size_t j_end = std::min(j_start + block_size, n); | |
| 83 | 72 | size_t k_start = broadcast_block * block_size; | |
| 84 | 72 | size_t k_end = std::min(k_start + block_size, n); | |
| 85 | |||
| 86 | 72 | ProcessBlock(matrix_a, matrix_b, output, n, i_start, i_end, j_start, j_end, k_start, k_end); | |
| 87 | } | ||
| 88 | } | ||
| 89 | } | ||
| 90 | 16 | } | |
| 91 | |||
| 92 | 160 | bool BaranovAMultMatrixFoxAlgorithmSEQ::RunImpl() { | |
| 93 | const auto &[matrix_size, matrix_a, matrix_b] = GetInput(); | ||
| 94 | 160 | size_t n = matrix_size; | |
| 95 | |||
| 96 | size_t block_size = 64; | ||
| 97 |
2/2✓ Branch 0 taken 144 times.
✓ Branch 1 taken 16 times.
|
160 | if (n < block_size) { |
| 98 | 144 | StandardMultiplication(n); | |
| 99 | } else { | ||
| 100 | 16 | FoxBlockMultiplication(n, block_size); | |
| 101 | } | ||
| 102 | |||
| 103 | 160 | return true; | |
| 104 | } | ||
| 105 | |||
| 106 | 160 | bool BaranovAMultMatrixFoxAlgorithmSEQ::PostProcessingImpl() { | |
| 107 | 160 | return true; | |
| 108 | } | ||
| 109 | |||
| 110 | } // namespace baranov_a_mult_matrix_fox_algorithm_seq | ||
| 111 |