| Line | Branch | Exec | Source |
|---|---|---|---|
| 1 | #include "chernov_t_radix_sort/stl/include/ops_stl.hpp" | ||
| 2 | |||
| 3 | #include <algorithm> | ||
| 4 | #include <cstddef> | ||
| 5 | #include <cstdint> | ||
| 6 | #include <thread> | ||
| 7 | #include <vector> | ||
| 8 | |||
| 9 | #include "chernov_t_radix_sort/common/include/common.hpp" | ||
| 10 | #include "util/include/util.hpp" | ||
| 11 | |||
| 12 | namespace chernov_t_radix_sort { | ||
| 13 | |||
| 14 |
1/2✓ Branch 1 taken 64 times.
✗ Branch 2 not taken.
|
64 | ChernovTRadixSortSTL::ChernovTRadixSortSTL(const InType &in) { |
| 15 | SetTypeOfTask(GetStaticTypeOfTask()); | ||
| 16 |
1/2✓ Branch 1 taken 64 times.
✗ Branch 2 not taken.
|
64 | GetInput() = in; |
| 17 | GetOutput() = {}; | ||
| 18 | 64 | } | |
| 19 | |||
| 20 | 64 | bool ChernovTRadixSortSTL::ValidationImpl() { | |
| 21 | 64 | return true; | |
| 22 | } | ||
| 23 | |||
| 24 | 64 | bool ChernovTRadixSortSTL::PreProcessingImpl() { | |
| 25 | 64 | GetOutput() = GetInput(); | |
| 26 | 64 | return true; | |
| 27 | } | ||
| 28 | |||
| 29 | constexpr int kBitsPerDigit = 8; | ||
| 30 | constexpr int kRadix = 1 << kBitsPerDigit; | ||
| 31 | constexpr uint32_t kSignMask = 0x80000000U; | ||
| 32 | |||
| 33 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 48 times.
|
48 | void ChernovTRadixSortSTL::RadixSortLSDSequential(std::vector<int> &data) { |
| 34 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 48 times.
|
48 | if (data.size() <= 1) { |
| 35 | ✗ | return; | |
| 36 | } | ||
| 37 | |||
| 38 | const size_t n = data.size(); | ||
| 39 | 48 | std::vector<uint32_t> temp(n); | |
| 40 |
2/2✓ Branch 0 taken 328 times.
✓ Branch 1 taken 48 times.
|
376 | for (size_t i = 0; i < n; ++i) { |
| 41 | 328 | temp[i] = static_cast<uint32_t>(data[i]) ^ kSignMask; | |
| 42 | } | ||
| 43 | |||
| 44 |
1/4✓ Branch 1 taken 48 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
|
48 | std::vector<uint32_t> buffer(n); |
| 45 | |||
| 46 |
2/2✓ Branch 0 taken 192 times.
✓ Branch 1 taken 48 times.
|
240 | for (int byte = 0; byte < 4; ++byte) { |
| 47 | 192 | const int shift = byte * kBitsPerDigit; | |
| 48 |
1/4✓ Branch 1 taken 192 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
|
192 | std::vector<int> count(kRadix, 0); |
| 49 | |||
| 50 |
2/2✓ Branch 0 taken 1312 times.
✓ Branch 1 taken 192 times.
|
1504 | for (size_t i = 0; i < n; ++i) { |
| 51 | 1312 | ++count[static_cast<int>((temp[i] >> shift) & 0xFFU)]; | |
| 52 | } | ||
| 53 |
2/2✓ Branch 0 taken 48960 times.
✓ Branch 1 taken 192 times.
|
49152 | for (int i = 1; i < kRadix; ++i) { |
| 54 | 48960 | count[i] += count[i - 1]; | |
| 55 | } | ||
| 56 |
2/2✓ Branch 0 taken 1312 times.
✓ Branch 1 taken 192 times.
|
1504 | for (size_t i = n; i-- > 0;) { |
| 57 | 1312 | buffer[static_cast<size_t>(--count[static_cast<int>((temp[i] >> shift) & 0xFFU)])] = temp[i]; | |
| 58 | } | ||
| 59 | temp.swap(buffer); | ||
| 60 | } | ||
| 61 | |||
| 62 |
2/2✓ Branch 0 taken 328 times.
✓ Branch 1 taken 48 times.
|
376 | for (size_t i = 0; i < n; ++i) { |
| 63 | 328 | data[i] = static_cast<int>(temp[i] ^ kSignMask); | |
| 64 | } | ||
| 65 | } | ||
| 66 | |||
| 67 | ✗ | void ChernovTRadixSortSTL::ConvertToIntegers(const std::vector<int> &input, std::vector<uint32_t> &output, size_t start, | |
| 68 | size_t end, uint32_t sign_mask) { | ||
| 69 | ✗ | for (size_t i = start; i < end; ++i) { | |
| 70 | ✗ | output[i] = static_cast<uint32_t>(input[i]) ^ sign_mask; | |
| 71 | } | ||
| 72 | ✗ | } | |
| 73 | |||
| 74 | ✗ | void ChernovTRadixSortSTL::ComputeLocalHistograms(const std::vector<uint32_t> &data, | |
| 75 | std::vector<std::vector<int>> &local_counts, size_t start, size_t end, | ||
| 76 | int shift, int thread_idx) { | ||
| 77 | ✗ | auto &cnt = local_counts[static_cast<size_t>(thread_idx)]; | |
| 78 | ✗ | for (size_t i = start; i < end; ++i) { | |
| 79 | ✗ | int digit = static_cast<int>((data[i] >> shift) & 0xFFU); | |
| 80 | ✗ | ++cnt[digit]; | |
| 81 | } | ||
| 82 | ✗ | } | |
| 83 | |||
| 84 | ✗ | void ChernovTRadixSortSTL::ComputeGlobalStarts(const std::vector<std::vector<int>> &local_counts, | |
| 85 | std::vector<int> &global_start, int k_radix, int num_threads) { | ||
| 86 | int current_pos = 0; | ||
| 87 | ✗ | for (int digit_idx = 0; digit_idx < k_radix; ++digit_idx) { | |
| 88 | int total = 0; | ||
| 89 | ✗ | for (int thread_idx = 0; thread_idx < num_threads; ++thread_idx) { | |
| 90 | ✗ | total += local_counts[static_cast<size_t>(thread_idx)][digit_idx]; | |
| 91 | } | ||
| 92 | ✗ | global_start[digit_idx] = current_pos; | |
| 93 | ✗ | current_pos += total; | |
| 94 | } | ||
| 95 | ✗ | } | |
| 96 | |||
| 97 | ✗ | void ChernovTRadixSortSTL::ComputeThreadOffsets(const std::vector<std::vector<int>> &local_counts, | |
| 98 | std::vector<std::vector<int>> &thread_offset, int k_radix, | ||
| 99 | int num_threads) { | ||
| 100 | ✗ | for (int digit_idx = 0; digit_idx < k_radix; ++digit_idx) { | |
| 101 | int offset = 0; | ||
| 102 | ✗ | for (int thread_idx = 0; thread_idx < num_threads; ++thread_idx) { | |
| 103 | ✗ | thread_offset[static_cast<size_t>(thread_idx)][digit_idx] = offset; | |
| 104 | ✗ | offset += local_counts[static_cast<size_t>(thread_idx)][digit_idx]; | |
| 105 | } | ||
| 106 | } | ||
| 107 | ✗ | } | |
| 108 | |||
| 109 | ✗ | void ChernovTRadixSortSTL::ScatterElements(const std::vector<uint32_t> &input, std::vector<uint32_t> &output, | |
| 110 | const std::vector<int> &global_start, | ||
| 111 | const std::vector<std::vector<int>> &thread_offset, | ||
| 112 | std::vector<std::vector<int>> &local_counter, size_t start, size_t end, | ||
| 113 | int shift, int thread_idx) { | ||
| 114 | ✗ | auto &my_counter = local_counter[static_cast<size_t>(thread_idx)]; | |
| 115 | ✗ | for (size_t i = start; i < end; ++i) { | |
| 116 | ✗ | int digit = static_cast<int>((input[i] >> shift) & 0xFFU); | |
| 117 | ✗ | int pos = global_start[digit] + thread_offset[static_cast<size_t>(thread_idx)][digit] + my_counter[digit]; | |
| 118 | ✗ | output[static_cast<size_t>(pos)] = input[i]; | |
| 119 | ✗ | ++my_counter[digit]; | |
| 120 | } | ||
| 121 | ✗ | } | |
| 122 | |||
| 123 | ✗ | void ChernovTRadixSortSTL::ConvertFromIntegers(const std::vector<uint32_t> &input, std::vector<int> &output, | |
| 124 | size_t start, size_t end, uint32_t sign_mask) { | ||
| 125 | ✗ | for (size_t i = start; i < end; ++i) { | |
| 126 | ✗ | output[i] = static_cast<int>(input[i] ^ sign_mask); | |
| 127 | } | ||
| 128 | ✗ | } | |
| 129 | |||
| 130 | ✗ | void ChernovTRadixSortSTL::RadixSortLSDParallel(std::vector<int> &data, int num_threads) { | |
| 131 | ✗ | const size_t n = data.size(); | |
| 132 | ✗ | if (n <= 1) { | |
| 133 | ✗ | return; | |
| 134 | } | ||
| 135 | |||
| 136 | ✗ | std::vector<uint32_t> temp(n); | |
| 137 | ✗ | std::vector<std::thread> threads; | |
| 138 | ✗ | const size_t chunk_size = (n + static_cast<size_t>(num_threads) - 1) / static_cast<size_t>(num_threads); | |
| 139 | |||
| 140 | ✗ | for (int thread_idx = 0; thread_idx < num_threads; ++thread_idx) { | |
| 141 | ✗ | const size_t start = static_cast<size_t>(thread_idx) * chunk_size; | |
| 142 | ✗ | const size_t end = std::min(start + chunk_size, n); | |
| 143 | ✗ | threads.emplace_back([&data, &temp, start, end]() { ConvertToIntegers(data, temp, start, end, kSignMask); }); | |
| 144 | } | ||
| 145 | ✗ | for (auto &th : threads) { | |
| 146 | ✗ | th.join(); | |
| 147 | } | ||
| 148 | ✗ | threads.clear(); | |
| 149 | |||
| 150 | ✗ | std::vector<uint32_t> buffer(n); | |
| 151 | |||
| 152 | ✗ | for (int byte_index = 0; byte_index < 4; ++byte_index) { | |
| 153 | ✗ | const int shift = byte_index * kBitsPerDigit; | |
| 154 | |||
| 155 | // Гистограммы | ||
| 156 | ✗ | std::vector<std::vector<int>> local_counts(static_cast<size_t>(num_threads), std::vector<int>(kRadix, 0)); | |
| 157 | ✗ | for (int thread_idx = 0; thread_idx < num_threads; ++thread_idx) { | |
| 158 | ✗ | const size_t start = static_cast<size_t>(thread_idx) * chunk_size; | |
| 159 | ✗ | const size_t end = std::min(start + chunk_size, n); | |
| 160 | ✗ | threads.emplace_back([&temp, &local_counts, start, end, shift, thread_idx]() { | |
| 161 | ✗ | ComputeLocalHistograms(temp, local_counts, start, end, shift, thread_idx); | |
| 162 | ✗ | }); | |
| 163 | } | ||
| 164 | ✗ | for (auto &th : threads) { | |
| 165 | ✗ | th.join(); | |
| 166 | } | ||
| 167 | ✗ | threads.clear(); | |
| 168 | |||
| 169 | ✗ | std::vector<int> global_start(kRadix, 0); | |
| 170 | ComputeGlobalStarts(local_counts, global_start, kRadix, num_threads); | ||
| 171 | |||
| 172 | ✗ | std::vector<std::vector<int>> thread_offset(static_cast<size_t>(num_threads), std::vector<int>(kRadix, 0)); | |
| 173 | ComputeThreadOffsets(local_counts, thread_offset, kRadix, num_threads); | ||
| 174 | |||
| 175 | ✗ | std::vector<std::vector<int>> local_counter(static_cast<size_t>(num_threads), std::vector<int>(kRadix, 0)); | |
| 176 | |||
| 177 | ✗ | for (int thread_idx = 0; thread_idx < num_threads; ++thread_idx) { | |
| 178 | ✗ | const size_t start = static_cast<size_t>(thread_idx) * chunk_size; | |
| 179 | ✗ | const size_t end = std::min(start + chunk_size, n); | |
| 180 | ✗ | threads.emplace_back( | |
| 181 | ✗ | [&temp, &buffer, &global_start, &thread_offset, &local_counter, start, end, shift, thread_idx]() { | |
| 182 | ✗ | ScatterElements(temp, buffer, global_start, thread_offset, local_counter, start, end, shift, thread_idx); | |
| 183 | ✗ | }); | |
| 184 | } | ||
| 185 | ✗ | for (auto &th : threads) { | |
| 186 | ✗ | th.join(); | |
| 187 | } | ||
| 188 | ✗ | threads.clear(); | |
| 189 | |||
| 190 | temp.swap(buffer); | ||
| 191 | ✗ | } | |
| 192 | |||
| 193 | ✗ | for (int thread_idx = 0; thread_idx < num_threads; ++thread_idx) { | |
| 194 | ✗ | const size_t start = static_cast<size_t>(thread_idx) * chunk_size; | |
| 195 | ✗ | const size_t end = std::min(start + chunk_size, n); | |
| 196 | ✗ | threads.emplace_back([&temp, &data, start, end]() { ConvertFromIntegers(temp, data, start, end, kSignMask); }); | |
| 197 | } | ||
| 198 | ✗ | for (auto &th : threads) { | |
| 199 | ✗ | th.join(); | |
| 200 | } | ||
| 201 | ✗ | } | |
| 202 | |||
| 203 |
4/4✓ Branch 0 taken 48 times.
✓ Branch 1 taken 16 times.
✓ Branch 2 taken 48 times.
✓ Branch 3 taken 16 times.
|
128 | bool ChernovTRadixSortSTL::RunImpl() { |
| 204 | auto &data = GetOutput(); | ||
| 205 |
2/2✓ Branch 0 taken 48 times.
✓ Branch 1 taken 16 times.
|
64 | if (data.size() <= 1) { |
| 206 | return true; | ||
| 207 | } | ||
| 208 | |||
| 209 |
1/2✓ Branch 0 taken 48 times.
✗ Branch 1 not taken.
|
48 | if (data.size() < 1000) { |
| 210 | 48 | RadixSortLSDSequential(data); | |
| 211 | 48 | return true; | |
| 212 | } | ||
| 213 | |||
| 214 | ✗ | const int num_threads = ppc::util::GetNumThreads(); | |
| 215 | ✗ | RadixSortLSDParallel(data, num_threads); | |
| 216 | ✗ | return true; | |
| 217 | } | ||
| 218 | |||
| 219 |
2/2✓ Branch 0 taken 56 times.
✓ Branch 1 taken 8 times.
|
64 | bool ChernovTRadixSortSTL::PostProcessingImpl() { |
| 220 | 64 | return std::is_sorted(GetOutput().begin(), GetOutput().end()); | |
| 221 | } | ||
| 222 | |||
| 223 | } // namespace chernov_t_radix_sort | ||
| 224 |