GCC Code Coverage Report

Directory: ./
File: tasks/kolotukhin_a_gaussian_blur/all/src/ops_all.cpp
Date: 2026-05-11 08:26:31
Exec Total Coverage
Lines: 112 118 94.9%
Functions: 12 12 100.0%
Branches: 58 86 67.4%
Line Branch Exec Source
1 #include "kolotukhin_a_gaussian_blur/all/include/ops_all.hpp"
2
3 #include <mpi.h>
4 #include <omp.h>
5
6 #include <algorithm>
7 #include <array>
8 #include <cstddef>
9 #include <cstdint>
10 #include <utility>
11 #include <vector>
12
13 #include "kolotukhin_a_gaussian_blur/common/include/common.hpp"
14
15 namespace kolotukhin_a_gaussian_blur {
16 namespace {
17 351 std::uint8_t GetPixel(const std::vector<std::uint8_t> &pixel_data, int img_width, int img_height, int pos_x,
18 int pos_y) {
19
4/4
✓ Branch 0 taken 36 times.
✓ Branch 1 taken 315 times.
✓ Branch 2 taken 243 times.
✓ Branch 3 taken 108 times.
 387 std::size_t x = static_cast<std::size_t>(std::max(0, std::min(pos_x, img_width - 1)));
20
4/4
✓ Branch 0 taken 39 times.
✓ Branch 1 taken 312 times.
✓ Branch 2 taken 195 times.
✓ Branch 3 taken 156 times.
 390 std::size_t y = static_cast<std::size_t>(std::max(0, std::min(pos_y, img_height - 1)));
21 351 return pixel_data[(y * static_cast<std::size_t>(img_width)) + x];
22 }
23
24 void ApplyGaussianBlur(const std::vector<std::uint8_t> &src_data, std::vector<std::uint8_t> &dst_data, int width,
25 int height, int start_row, int end_row) {
26 const static std::array<std::array<int, 3>, 3> kKernel = {{{{1, 2, 1}}, {{2, 4, 2}}, {{1, 2, 1}}}};
27 const static int kSum = 16;
28 8 #pragma omp parallel for collapse(2) schedule(static) default(none) \
29 shared(src_data, dst_data, width, height, start_row, end_row, kKernel, kSum)
30 for (int row = start_row; row < end_row; row++) {
31 for (int col = 0; col < width; col++) {
32 int acc = 0;
33 for (int dy = -1; dy <= 1; dy++) {
34 for (int dx = -1; dx <= 1; dx++) {
35 std::uint8_t pixel = GetPixel(src_data, width, height, col + dx, row + dy);
36 acc += kKernel.at(1 + dy).at(1 + dx) * static_cast<int>(pixel);
37 }
38 }
39 dst_data[(static_cast<std::size_t>(row) * static_cast<std::size_t>(width)) + static_cast<std::size_t>(col)] =
40 static_cast<std::uint8_t>(acc / kSum);
41 }
42 }
43 }
44 } // namespace
45
46 4 void KolotukhinAGaussinBlurALL::SendWorkData(int rows_per_process, int remainder) {
47 const auto &pixel_data = get<0>(GetInput());
48 int current_row = 0;
49
50
2/2
✓ Branch 0 taken 8 times.
✓ Branch 1 taken 4 times.
 12 for (int dest = 0; dest < proc_count_; dest++) {
51
2/2
✓ Branch 0 taken 4 times.
✓ Branch 1 taken 4 times.
 8 int dest_rows = (dest < remainder) ? rows_per_process + 1 : rows_per_process;
52
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 8 times.
 8 if (dest_rows == 0) {
53 ✗ continue;
54 }
55 int start_row = current_row;
56 8 int end_row = current_row + dest_rows;
57
58 8 int extended_start = std::max(0, start_row - 1);
59 8 int extended_end = std::min(global_height_, end_row + 1);
60 8 int extended_rows = extended_end - extended_start;
61
62 8 std::vector<std::uint8_t> extended_data(static_cast<std::size_t>(extended_rows) *
63 8 static_cast<std::size_t>(global_width_));
64
65 16 std::copy(pixel_data.begin() + static_cast<std::ptrdiff_t>(extended_start) * global_width_,
66 8 pixel_data.begin() + static_cast<std::ptrdiff_t>(extended_end) * global_width_, extended_data.begin());
67
68
2/2
✓ Branch 0 taken 4 times.
✓ Branch 1 taken 4 times.
 8 if (dest == 0) {
69 4 local_data_ = std::move(extended_data);
70 } else {
71
1/2
✓ Branch 1 taken 4 times.
✗ Branch 2 not taken.
 4 MPI_Send(extended_data.data(), static_cast<int>(extended_data.size()), MPI_UNSIGNED_CHAR, dest, 0,
72 MPI_COMM_WORLD);
73 }
74 current_row += dest_rows;
75 }
76 4 }
77
78 4 void KolotukhinAGaussinBlurALL::ReceiveWorkData() {
79 4 MPI_Recv(local_data_.data(), static_cast<int>(local_data_.size()), MPI_UNSIGNED_CHAR, 0, 0, MPI_COMM_WORLD,
80 MPI_STATUS_IGNORE);
81 4 }
82
83 8 void KolotukhinAGaussinBlurALL::DistributeWork() {
84 8 int rows_per_process = global_height_ / proc_count_;
85 8 int remainder = global_height_ % proc_count_;
86
87 8 int last_handler = proc_count_ - 1;
88
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 8 times.
 8 if (rows_per_process == 0) {
89 ✗ last_handler = remainder - 1;
90 }
91
92
2/2
✓ Branch 0 taken 4 times.
✓ Branch 1 taken 4 times.
 8 local_height_ = (rank_ < remainder) ? rows_per_process + 1 : rows_per_process;
93
94
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 8 times.
 8 if (local_height_ == 0) {
95 local_data_.clear();
96 ✗ return;
97 }
98
99
3/4
✓ Branch 0 taken 4 times.
✓ Branch 1 taken 4 times.
✓ Branch 2 taken 4 times.
✗ Branch 3 not taken.
 8 local_height_ = (rank_ == 0 || rank_ == last_handler) ? local_height_ + 1 : local_height_ + 2;
100 8 std::size_t local_size = static_cast<std::size_t>(local_height_) * static_cast<std::size_t>(global_width_);
101 8 local_data_.resize(local_size, 0);
102
103
2/2
✓ Branch 0 taken 4 times.
✓ Branch 1 taken 4 times.
 8 if (rank_ == 0) {
104 4 SendWorkData(rows_per_process, remainder);
105 } else {
106 4 ReceiveWorkData();
107 }
108 }
109
110
1/2
✓ Branch 1 taken 8 times.
✗ Branch 2 not taken.
 8 KolotukhinAGaussinBlurALL::KolotukhinAGaussinBlurALL(const InType &in) {
111 SetTypeOfTask(GetStaticTypeOfTask());
112 GetInput() = in;
113 GetOutput().clear();
114
115
1/2
✓ Branch 1 taken 8 times.
✗ Branch 2 not taken.
 8 MPI_Comm_rank(MPI_COMM_WORLD, &rank_);
116
1/2
✓ Branch 1 taken 8 times.
✗ Branch 2 not taken.
 8 MPI_Comm_size(MPI_COMM_WORLD, &proc_count_);
117 8 }
118
119 8 bool KolotukhinAGaussinBlurALL::ValidationImpl() {
120 const auto &pixel_data = get<0>(GetInput());
121 8 const auto img_width = get<1>(GetInput());
122 8 const auto img_height = get<2>(GetInput());
123
124
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 8 times.
 8 bool valid = static_cast<std::size_t>(img_height) * static_cast<std::size_t>(img_width) == pixel_data.size();
125
126
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 8 times.
 8 int local_valid = valid ? 1 : 0;
127 8 int global_valid = 0;
128 8 MPI_Allreduce(&local_valid, &global_valid, 1, MPI_INT, MPI_MIN, MPI_COMM_WORLD);
129 8 return global_valid == 1;
130 }
131
132 8 bool KolotukhinAGaussinBlurALL::PreProcessingImpl() {
133 8 const auto img_width = get<1>(GetInput());
134 8 const auto img_height = get<2>(GetInput());
135
136 8 int width = img_width;
137 8 int height = img_height;
138
139 8 MPI_Bcast(&width, 1, MPI_INT, 0, MPI_COMM_WORLD);
140 8 MPI_Bcast(&height, 1, MPI_INT, 0, MPI_COMM_WORLD);
141
142
2/2
✓ Branch 0 taken 4 times.
✓ Branch 1 taken 4 times.
 8 if (rank_ != 0) {
143 4 global_width_ = width;
144 4 global_height_ = height;
145 } else {
146 4 global_width_ = img_width;
147 4 global_height_ = img_height;
148 }
149
150
2/2
✓ Branch 0 taken 4 times.
✓ Branch 1 taken 4 times.
 8 if (rank_ == 0) {
151 4 GetOutput().assign(static_cast<std::size_t>(global_height_) * static_cast<std::size_t>(global_width_), 0);
152 }
153
154 8 DistributeWork();
155 8 return true;
156 }
157
158 8 void KolotukhinAGaussinBlurALL::GatherResults() {
159 8 int rows_per_process = global_height_ / proc_count_;
160 8 int remainder = global_height_ % proc_count_;
161
162
2/2
✓ Branch 0 taken 4 times.
✓ Branch 1 taken 4 times.
 8 if (rank_ != 0) {
163 4 GatherResultsWorker(rows_per_process, remainder);
164 } else {
165 4 GatherResultsRoot(rows_per_process, remainder);
166 }
167 8 }
168
169 4 void KolotukhinAGaussinBlurALL::GatherResultsWorker(int rows_per_process, int remainder) {
170
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 4 times.
 4 int original_rows = (rank_ < remainder) ? rows_per_process + 1 : rows_per_process;
171
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 4 times.
 4 if (original_rows <= 0) {
172 ✗ return;
173 }
174
175 4 int halo_offset = (rank_ == 0) ? 0 : 1;
176 4 std::vector<std::uint8_t> result_only_own(static_cast<std::size_t>(original_rows) *
177 4 static_cast<std::size_t>(global_width_));
178
179
1/2
✓ Branch 1 taken 4 times.
✗ Branch 2 not taken.
 4 std::copy(local_data_.begin() + static_cast<std::ptrdiff_t>(halo_offset) * global_width_,
180 4 local_data_.begin() + static_cast<std::ptrdiff_t>(halo_offset + original_rows) * global_width_,
181 result_only_own.begin());
182
183
1/2
✓ Branch 1 taken 4 times.
✗ Branch 2 not taken.
 4 MPI_Send(result_only_own.data(), static_cast<int>(result_only_own.size()), MPI_UNSIGNED_CHAR, 0, 1, MPI_COMM_WORLD);
184 }
185
186 4 void KolotukhinAGaussinBlurALL::GatherResultsRoot(int rows_per_process, int remainder) {
187 auto &output = GetOutput();
188 4 std::vector<int> recv_counts(proc_count_);
189
1/4
✓ Branch 1 taken 4 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
 4 std::vector<int> displs(proc_count_);
190
191 int current_row = 0;
192
2/2
✓ Branch 0 taken 8 times.
✓ Branch 1 taken 4 times.
 12 for (int i = 0; i < proc_count_; i++) {
193
2/2
✓ Branch 0 taken 4 times.
✓ Branch 1 taken 4 times.
 8 int rows = (i < remainder) ? rows_per_process + 1 : rows_per_process;
194 8 recv_counts[i] = rows * global_width_;
195 8 displs[i] = current_row * global_width_;
196 8 current_row += rows;
197 }
198
199
1/2
✓ Branch 0 taken 4 times.
✗ Branch 1 not taken.
 4 int root_original_rows = (0 < remainder) ? rows_per_process + 1 : rows_per_process;
200 4 std::copy(local_data_.begin(), local_data_.begin() + static_cast<std::ptrdiff_t>(root_original_rows) * global_width_,
201 output.begin() + displs[0]);
202
203
2/2
✓ Branch 0 taken 4 times.
✓ Branch 1 taken 4 times.
 8 for (int src = 1; src < proc_count_; src++) {
204
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 4 times.
 4 int src_rows = (src < remainder) ? rows_per_process + 1 : rows_per_process;
205
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 4 times.
 4 if (src_rows == 0) {
206 ✗ continue;
207 }
208
209
2/6
✓ Branch 1 taken 4 times.
✗ Branch 2 not taken.
✓ Branch 4 taken 4 times.
✗ Branch 5 not taken.
✗ Branch 6 not taken.
✗ Branch 7 not taken.
 4 std::vector<std::uint8_t> src_data(static_cast<std::size_t>(src_rows) * static_cast<std::size_t>(global_width_));
210
1/2
✓ Branch 1 taken 4 times.
✗ Branch 2 not taken.
 4 MPI_Recv(src_data.data(), static_cast<int>(src_data.size()), MPI_UNSIGNED_CHAR, src, 1, MPI_COMM_WORLD,
211 MPI_STATUS_IGNORE);
212
213
1/2
✓ Branch 0 taken 4 times.
✗ Branch 1 not taken.
 4 std::ranges::copy(src_data, output.begin() + displs[src]);
214 }
215 4 }
216
217 8 bool KolotukhinAGaussinBlurALL::RunImpl() {
218
1/2
✓ Branch 0 taken 8 times.
✗ Branch 1 not taken.
 8 if (local_height_ == 0) {
219 return true;
220 }
221
222 8 int rows_per_process = global_height_ / proc_count_;
223 8 int remainder = global_height_ % proc_count_;
224
225 8 int last_handler = proc_count_ - 1;
226
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 8 times.
 8 if (rows_per_process == 0) {
227 ✗ last_handler = remainder - 1;
228 }
229
230 8 int extended_height = static_cast<int>(local_data_.size() / global_width_);
231
232 8 int begin = (rank_ == 0) ? 0 : 1;
233
2/2
✓ Branch 0 taken 4 times.
✓ Branch 1 taken 4 times.
 8 int end = (rank_ == last_handler) ? extended_height : extended_height - 1;
234
235 8 std::vector<std::uint8_t> result(local_data_.size());
236 8 ApplyGaussianBlur(local_data_, result, global_width_, extended_height, begin, end);
237
238 local_data_ = std::move(result);
239 return true;
240 }
241
242 8 bool KolotukhinAGaussinBlurALL::PostProcessingImpl() {
243 8 GatherResults();
244 8 return true;
245 }
246
247 } // namespace kolotukhin_a_gaussian_blur
248