GCC Code Coverage Report

Directory: ./
File: tasks/zagryadskov_m_complex_spmm_ccs/all/src/ops_all.cpp
Date: 2026-05-11 08:26:31
Exec Total Coverage
Lines: 200 202 99.0%
Functions: 14 14 100.0%
Branches: 103 150 68.7%
Line Branch Exec Source
1 #include "zagryadskov_m_complex_spmm_ccs/all/include/ops_all.hpp"
2
3 #include <mpi.h>
4 #include <omp.h>
5
6 #include <algorithm>
7 #include <complex>
8 #include <stdexcept>
9 #include <tuple>
10 #include <vector>
11
12 #include "util/include/util.hpp"
13 #include "zagryadskov_m_complex_spmm_ccs/common/include/common.hpp"
14
15 namespace zagryadskov_m_complex_spmm_ccs {
16
17
1/2
✓ Branch 2 taken 6 times.
✗ Branch 3 not taken.
 6 ZagryadskovMComplexSpMMCCSALL::ZagryadskovMComplexSpMMCCSALL(const InType &in) {
18 SetTypeOfTask(GetStaticTypeOfTask());
19 6 int world_rank = 0;
20 int err_code = 0;
21
1/2
✓ Branch 1 taken 6 times.
✗ Branch 2 not taken.
 6 err_code = MPI_Comm_rank(MPI_COMM_WORLD, &world_rank);
22
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 6 times.
 6 if (err_code != MPI_SUCCESS) {
23 ✗ throw std::runtime_error("MPI_Comm_rank failed");
24 }
25
2/2
✓ Branch 0 taken 3 times.
✓ Branch 1 taken 3 times.
 6 if (world_rank == 0) {
26 GetInput() = in;
27
1/2
✓ Branch 1 taken 3 times.
✗ Branch 2 not taken.
 3 GetOutput() = CCS();
28 } else {
29
1/2
✓ Branch 1 taken 3 times.
✗ Branch 2 not taken.
 6 GetInput() = std::make_tuple(CCS(), CCS());
30
1/2
✓ Branch 1 taken 3 times.
✗ Branch 2 not taken.
 3 GetOutput() = CCS();
31 }
32 6 }
33
34 12 void ZagryadskovMComplexSpMMCCSALL::SpMMSymbolic(const CCS &a, const CCS &b, std::vector<int> &col_ptr, int jstart,
35 int jend) {
36 12 std::vector<int> marker(a.m, -1);
37
38
2/2
✓ Branch 0 taken 7 times.
✓ Branch 1 taken 12 times.
 19 for (int j = jstart; j < jend; ++j) {
39 int count = 0;
40
41
2/2
✓ Branch 0 taken 10 times.
✓ Branch 1 taken 7 times.
 17 for (int k = b.col_ptr[j]; k < b.col_ptr[j + 1]; ++k) {
42 10 int b_row = b.row_ind[k];
43
2/2
✓ Branch 0 taken 11 times.
✓ Branch 1 taken 10 times.
 21 for (int zp = a.col_ptr[b_row]; zp < a.col_ptr[b_row + 1]; ++zp) {
44
2/2
✓ Branch 0 taken 9 times.
✓ Branch 1 taken 2 times.
 11 int a_row = a.row_ind[zp];
45
2/2
✓ Branch 0 taken 9 times.
✓ Branch 1 taken 2 times.
 11 if (marker[a_row] != j) {
46 9 marker[a_row] = j;
47 9 ++count;
48 }
49 }
50 }
51 7 col_ptr[j + 1] += count;
52 }
53 12 }
54
55
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 7 times.
 7 void ZagryadskovMComplexSpMMCCSALL::SpMMKernel(const CCS &a, const CCS &b, CCS &c, const std::complex<double> &zero,
56 std::vector<int> &rows, std::vector<std::complex<double>> &acc,
57 std::vector<int> &marker, int j) {
58 rows.clear();
59 7 int write_ptr = c.col_ptr[j];
60
61
2/2
✓ Branch 0 taken 10 times.
✓ Branch 1 taken 7 times.
 17 for (int k = b.col_ptr[j]; k < b.col_ptr[j + 1]; ++k) {
62 10 std::complex<double> tmpval = b.values[k];
63 10 int b_row = b.row_ind[k];
64
2/2
✓ Branch 0 taken 11 times.
✓ Branch 1 taken 10 times.
 21 for (int zp = a.col_ptr[b_row]; zp < a.col_ptr[b_row + 1]; ++zp) {
65
2/2
✓ Branch 0 taken 9 times.
✓ Branch 1 taken 2 times.
 11 int a_row = a.row_ind[zp];
66
2/2
✓ Branch 0 taken 9 times.
✓ Branch 1 taken 2 times.
 11 acc[a_row] += tmpval * a.values[zp];
67
2/2
✓ Branch 0 taken 9 times.
✓ Branch 1 taken 2 times.
 11 if (marker[a_row] != j) {
68
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 9 times.
 9 marker[a_row] = j;
69 rows.push_back(a_row);
70 }
71 }
72 }
73
74
2/2
✓ Branch 0 taken 9 times.
✓ Branch 1 taken 7 times.
 16 for (int r_idx : rows) {
75 9 c.row_ind[write_ptr] = r_idx;
76 9 c.values[write_ptr] = acc[r_idx];
77 9 ++write_ptr;
78 9 acc[r_idx] = zero;
79 }
80 7 }
81
82 12 void ZagryadskovMComplexSpMMCCSALL::SpMMNumeric(const CCS &a, const CCS &b, CCS &c, const std::complex<double> &zero,
83 int jstart, int jend) {
84 12 std::vector<int> marker(a.m, -1);
85
1/4
✓ Branch 1 taken 12 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
 12 std::vector<std::complex<double>> acc(a.m, zero);
86 12 std::vector<int> rows;
87
88
2/2
✓ Branch 0 taken 7 times.
✓ Branch 1 taken 12 times.
 19 for (int j = jstart; j < jend; ++j) {
89
1/2
✓ Branch 1 taken 7 times.
✗ Branch 2 not taken.
 7 SpMMKernel(a, b, c, zero, rows, acc, marker, j);
90 }
91 12 }
92
93 6 void ZagryadskovMComplexSpMMCCSALL::SpMM(const CCS &a, const CCS &b, CCS &c) {
94 6 c.m = a.m;
95 6 c.n = b.n;
96 6 const int num_threads = ppc::util::GetNumThreads();
97
98 6 std::complex<double> zero(0.0, 0.0);
99 6 c.col_ptr.assign(c.n + 1, 0);
100
101 6 #pragma omp parallel default(none) shared(num_threads, a, b, c) num_threads(ppc::util::GetNumThreads())
102 {
103 int tid = omp_get_thread_num();
104 int jstart = (tid * b.n) / num_threads;
105 int jend = ((tid + 1) * b.n) / num_threads;
106 SpMMSymbolic(a, b, c.col_ptr, jstart, jend);
107 }
108
109
2/2
✓ Branch 0 taken 7 times.
✓ Branch 1 taken 6 times.
 13 for (int j = 0; j < c.n; ++j) {
110 7 c.col_ptr[j + 1] += c.col_ptr[j];
111 }
112 6 int nnz = c.col_ptr[b.n];
113 6 c.row_ind.resize(nnz);
114 6 c.values.resize(nnz);
115 6 #pragma omp parallel default(none) shared(num_threads, a, b, c, zero) num_threads(ppc::util::GetNumThreads())
116 {
117 int tid = omp_get_thread_num();
118 int jstart = (tid * b.n) / num_threads;
119 int jend = ((tid + 1) * b.n) / num_threads;
120 SpMMNumeric(a, b, c, zero, jstart, jend);
121 }
122 6 }
123
124 6 void ZagryadskovMComplexSpMMCCSALL::BcastCCS(CCS &a, int rank) {
125 6 MPI_Bcast(&a.m, 1, MPI_INT, 0, MPI_COMM_WORLD);
126 6 MPI_Bcast(&a.n, 1, MPI_INT, 0, MPI_COMM_WORLD);
127
128 6 int nz = 0;
129
2/2
✓ Branch 0 taken 3 times.
✓ Branch 1 taken 3 times.
 6 if (rank == 0) {
130 3 nz = static_cast<int>(a.values.size());
131 }
132 6 MPI_Bcast(&nz, 1, MPI_INT, 0, MPI_COMM_WORLD);
133
134
2/2
✓ Branch 0 taken 3 times.
✓ Branch 1 taken 3 times.
 6 if (rank != 0) {
135 3 a.col_ptr.resize(a.n + 1);
136 3 a.row_ind.resize(nz);
137 3 a.values.resize(nz);
138 }
139
140 6 MPI_Bcast(a.col_ptr.data(), a.n + 1, MPI_INT, 0, MPI_COMM_WORLD);
141 6 MPI_Bcast(a.row_ind.data(), nz, MPI_INT, 0, MPI_COMM_WORLD);
142 6 MPI_Bcast(a.values.data(), nz, MPI_C_DOUBLE_COMPLEX, 0, MPI_COMM_WORLD);
143 6 }
144
145 3 void ZagryadskovMComplexSpMMCCSALL::SendCCS(const CCS &m, int dest) {
146 3 MPI_Send(&m.m, 1, MPI_INT, dest, 0, MPI_COMM_WORLD);
147 3 MPI_Send(&m.n, 1, MPI_INT, dest, 0, MPI_COMM_WORLD);
148 3 int nz = static_cast<int>(m.values.size());
149 3 MPI_Send(&nz, 1, MPI_INT, dest, 0, MPI_COMM_WORLD);
150
151 3 MPI_Send(m.col_ptr.data(), m.n + 1, MPI_INT, dest, 0, MPI_COMM_WORLD);
152 3 MPI_Send(m.row_ind.data(), nz, MPI_INT, dest, 0, MPI_COMM_WORLD);
153 3 MPI_Send(m.values.data(), nz, MPI_C_DOUBLE_COMPLEX, dest, 0, MPI_COMM_WORLD);
154 3 }
155
156 3 void ZagryadskovMComplexSpMMCCSALL::RecvCCS(CCS &m, int src) {
157 MPI_Status st;
158 3 MPI_Recv(&m.m, 1, MPI_INT, src, 0, MPI_COMM_WORLD, &st);
159 3 MPI_Recv(&m.n, 1, MPI_INT, src, 0, MPI_COMM_WORLD, &st);
160 3 int nz = 0;
161 3 MPI_Recv(&nz, 1, MPI_INT, src, 0, MPI_COMM_WORLD, &st);
162
163 3 m.col_ptr.resize(m.n + 1);
164 3 m.row_ind.resize(nz);
165 3 m.values.resize(nz);
166
167 3 MPI_Recv(m.col_ptr.data(), m.n + 1, MPI_INT, src, 0, MPI_COMM_WORLD, &st);
168 3 MPI_Recv(m.row_ind.data(), nz, MPI_INT, src, 0, MPI_COMM_WORLD, &st);
169 3 MPI_Recv(m.values.data(), nz, MPI_C_DOUBLE_COMPLEX, src, 0, MPI_COMM_WORLD, &st);
170 3 }
171
172 6 void ZagryadskovMComplexSpMMCCSALL::ScatterB(const CCS &b, CCS &b_local, const std::vector<int> &col_starts, int rank,
173 int size) {
174
2/2
✓ Branch 0 taken 3 times.
✓ Branch 1 taken 3 times.
 6 if (rank == 0) {
175 3 CCS tmp;
176
2/2
✓ Branch 0 taken 6 times.
✓ Branch 1 taken 3 times.
 9 for (int proc = 0; proc < size; ++proc) {
177
2/2
✓ Branch 0 taken 3 times.
✓ Branch 1 taken 3 times.
 6 int jstart = col_starts[proc];
178 6 int jend = col_starts[proc + 1];
179
180 6 tmp.m = b.m;
181
2/2
✓ Branch 0 taken 3 times.
✓ Branch 1 taken 3 times.
 6 tmp.n = jend - jstart;
182 tmp.row_ind.clear();
183 tmp.values.clear();
184 tmp.col_ptr.clear();
185
186
1/2
✓ Branch 1 taken 6 times.
✗ Branch 2 not taken.
 6 int nnz_start = b.col_ptr[jstart];
187
1/2
✓ Branch 1 taken 6 times.
✗ Branch 2 not taken.
 6 int nnz_end = b.col_ptr[jend];
188
1/2
✓ Branch 1 taken 6 times.
✗ Branch 2 not taken.
 6 tmp.row_ind.assign(b.row_ind.begin() + nnz_start, b.row_ind.begin() + nnz_end);
189 6 tmp.values.assign(b.values.begin() + nnz_start, b.values.begin() + nnz_end);
190
1/2
✓ Branch 1 taken 6 times.
✗ Branch 2 not taken.
 6 tmp.col_ptr.resize(tmp.n + 1);
191
2/2
✓ Branch 0 taken 13 times.
✓ Branch 1 taken 6 times.
 19 for (int j = 0; j <= tmp.n; ++j) {
192 13 tmp.col_ptr[j] = b.col_ptr[jstart + j] - nnz_start;
193 }
194
195
2/2
✓ Branch 0 taken 3 times.
✓ Branch 1 taken 3 times.
 6 if (proc == 0) {
196
1/2
✓ Branch 1 taken 3 times.
✗ Branch 2 not taken.
 3 b_local = tmp;
197 } else {
198
1/2
✓ Branch 1 taken 3 times.
✗ Branch 2 not taken.
 3 SendCCS(tmp, proc);
199 }
200 }
201 3 } else {
202 3 RecvCCS(b_local, 0);
203 }
204 6 }
205
206
1/2
✓ Branch 1 taken 6 times.
✗ Branch 2 not taken.
 6 void ZagryadskovMComplexSpMMCCSALL::GatherC(CCS &c, CCS &c_local, int rank, int size) {
207 MPI_Status st;
208 6 int local_nnz = static_cast<int>(c_local.values.size());
209 int total_nnz = 0;
210 6 int local_cols = c_local.n;
211 int total_cols = 0;
212 6 std::vector<int> tmp;
213
1/4
✓ Branch 1 taken 6 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
 6 std::vector<int> recvcounts(size);
214
2/6
✓ Branch 1 taken 6 times.
✗ Branch 2 not taken.
✓ Branch 4 taken 6 times.
✗ Branch 5 not taken.
✗ Branch 6 not taken.
✗ Branch 7 not taken.
 6 std::vector<int> displs(size);
215
216
1/2
✓ Branch 1 taken 6 times.
✗ Branch 2 not taken.
 6 MPI_Gather(&local_nnz, 1, MPI_INT, recvcounts.data(), 1, MPI_INT, 0, MPI_COMM_WORLD);
217
218
2/2
✓ Branch 0 taken 3 times.
✓ Branch 1 taken 3 times.
 6 if (rank == 0) {
219 3 c.m = c_local.m;
220
2/2
✓ Branch 0 taken 6 times.
✓ Branch 1 taken 3 times.
 9 for (int i = 0; i < size; ++i) {
221 6 displs[i] = total_nnz;
222 6 total_nnz += recvcounts[i];
223 }
224
1/2
✓ Branch 1 taken 3 times.
✗ Branch 2 not taken.
 3 c.row_ind.resize(total_nnz);
225
1/2
✓ Branch 1 taken 3 times.
✗ Branch 2 not taken.
 3 c.values.resize(total_nnz);
226 }
227
228
1/2
✓ Branch 1 taken 6 times.
✗ Branch 2 not taken.
 6 MPI_Gatherv(c_local.row_ind.data(), local_nnz, MPI_INT, c.row_ind.data(), recvcounts.data(), displs.data(), MPI_INT,
229 0, MPI_COMM_WORLD);
230
1/2
✓ Branch 1 taken 6 times.
✗ Branch 2 not taken.
 6 MPI_Gatherv(c_local.values.data(), local_nnz, MPI_C_DOUBLE_COMPLEX, c.values.data(), recvcounts.data(), displs.data(),
231 MPI_C_DOUBLE_COMPLEX, 0, MPI_COMM_WORLD);
232
233
1/2
✓ Branch 1 taken 6 times.
✗ Branch 2 not taken.
 6 MPI_Gather(&local_cols, 1, MPI_INT, recvcounts.data(), 1, MPI_INT, 0, MPI_COMM_WORLD);
234
235
2/2
✓ Branch 0 taken 3 times.
✓ Branch 1 taken 3 times.
 6 if (rank == 0) {
236
2/2
✓ Branch 0 taken 6 times.
✓ Branch 1 taken 3 times.
 9 for (int i = 0; i < size; ++i) {
237 6 displs[i] = total_cols + 1;
238 6 total_cols += recvcounts[i];
239 6 recvcounts[i] += 1;
240 }
241 3 c.n = total_cols;
242
1/2
✓ Branch 1 taken 3 times.
✗ Branch 2 not taken.
 3 c.col_ptr.resize(total_cols + 1);
243 }
244
245 if (rank != 0) {
246
1/2
✓ Branch 1 taken 3 times.
✗ Branch 2 not taken.
 3 MPI_Send(c_local.col_ptr.data(), c_local.n + 1, MPI_INT, 0, 0, MPI_COMM_WORLD);
247 }
248
249 if (rank == 0) {
250 std::ranges::copy(c_local.col_ptr, c.col_ptr.begin());
251
252 3 int nz_offset = c_local.col_ptr.back();
253 3 int col_offset = c_local.n;
254
2/2
✓ Branch 0 taken 3 times.
✓ Branch 1 taken 3 times.
 6 for (int proc = 1; proc < size; ++proc) {
255
1/2
✓ Branch 1 taken 3 times.
✗ Branch 2 not taken.
 3 tmp.resize(recvcounts[proc]);
256
1/2
✓ Branch 1 taken 3 times.
✗ Branch 2 not taken.
 3 MPI_Recv(tmp.data(), recvcounts[proc], MPI_INT, proc, 0, MPI_COMM_WORLD, &st);
257
258
2/2
✓ Branch 0 taken 4 times.
✓ Branch 1 taken 3 times.
 7 for (int j = 1; j < recvcounts[proc]; ++j) {
259 4 c.col_ptr[col_offset + j] = nz_offset + tmp[j];
260 }
261
262 3 nz_offset += tmp.back();
263
1/2
✓ Branch 0 taken 3 times.
✗ Branch 1 not taken.
 3 col_offset += recvcounts[proc] - 1;
264 tmp.clear();
265 }
266 }
267 6 }
268
269 6 bool ZagryadskovMComplexSpMMCCSALL::ValidationImpl() {
270 bool res = false;
271 6 int world_rank = 0;
272 int err_code = 0;
273 6 err_code = MPI_Comm_rank(MPI_COMM_WORLD, &world_rank);
274
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 6 times.
 6 if (err_code != MPI_SUCCESS) {
275 ✗ throw std::runtime_error("MPI_Comm_rank failed");
276 }
277
2/2
✓ Branch 0 taken 3 times.
✓ Branch 1 taken 3 times.
 6 if (world_rank == 0) {
278 const CCS &a = std::get<0>(GetInput());
279 const CCS &b = std::get<1>(GetInput());
280 3 res = a.n == b.m;
281 } else {
282 res = true;
283 }
284 6 return res;
285 }
286
287 6 bool ZagryadskovMComplexSpMMCCSALL::PreProcessingImpl() {
288 6 return true;
289 }
290
291 6 bool ZagryadskovMComplexSpMMCCSALL::RunImpl() {
292 6 int world_rank = 0;
293 6 int world_size = 0;
294 6 MPI_Comm_rank(MPI_COMM_WORLD, &world_rank);
295 6 MPI_Comm_size(MPI_COMM_WORLD, &world_size);
296 CCS &a = std::get<0>(GetInput());
297 CCS &b = std::get<1>(GetInput());
298 CCS &c = GetOutput();
299
300 6 CCS local_b;
301 6 CCS local_c;
302 6 std::vector<int> col_starts;
303
2/2
✓ Branch 0 taken 3 times.
✓ Branch 1 taken 3 times.
 6 if (world_rank == 0) {
304
1/2
✓ Branch 1 taken 3 times.
✗ Branch 2 not taken.
 3 col_starts.resize(world_size + 1);
305
2/2
✓ Branch 0 taken 9 times.
✓ Branch 1 taken 3 times.
 12 for (int proc = 0; proc <= world_size; ++proc) {
306 9 col_starts[proc] = (proc * b.n) / world_size;
307 }
308 }
309
310
1/2
✓ Branch 1 taken 6 times.
✗ Branch 2 not taken.
 6 BcastCCS(a, world_rank);
311
1/2
✓ Branch 1 taken 6 times.
✗ Branch 2 not taken.
 6 ScatterB(b, local_b, col_starts, world_rank, world_size);
312
313
1/2
✓ Branch 1 taken 6 times.
✗ Branch 2 not taken.
 6 ZagryadskovMComplexSpMMCCSALL::SpMM(a, local_b, local_c);
314
315
1/2
✓ Branch 1 taken 6 times.
✗ Branch 2 not taken.
 6 GatherC(c, local_c, world_rank, world_size);
316
317 6 return true;
318 6 }
319
320 6 bool ZagryadskovMComplexSpMMCCSALL::PostProcessingImpl() {
321 6 int world_rank = 0;
322 6 MPI_Comm_rank(MPI_COMM_WORLD, &world_rank);
323 6 int m = 0;
324 6 int n = 0;
325 6 int nz = 0;
326 CCS &c = GetOutput();
327
2/2
✓ Branch 0 taken 3 times.
✓ Branch 1 taken 3 times.
 6 if (world_rank == 0) {
328 3 m = c.m;
329 3 n = c.n;
330 3 nz = static_cast<int>(c.values.size());
331 }
332 6 MPI_Bcast(&m, 1, MPI_INT, 0, MPI_COMM_WORLD);
333 6 MPI_Bcast(&n, 1, MPI_INT, 0, MPI_COMM_WORLD);
334 6 MPI_Bcast(&nz, 1, MPI_INT, 0, MPI_COMM_WORLD);
335
2/2
✓ Branch 0 taken 3 times.
✓ Branch 1 taken 3 times.
 6 if (world_rank != 0) {
336 3 c.m = m;
337 3 c.n = n;
338 3 c.col_ptr.resize(n + 1);
339 3 c.row_ind.resize(nz);
340 3 c.values.resize(nz);
341 }
342 6 MPI_Bcast(c.col_ptr.data(), n + 1, MPI_INT, 0, MPI_COMM_WORLD);
343 6 MPI_Bcast(c.row_ind.data(), nz, MPI_INT, 0, MPI_COMM_WORLD);
344 6 MPI_Bcast(c.values.data(), nz, MPI_C_DOUBLE_COMPLEX, 0, MPI_COMM_WORLD);
345
346 6 return true;
347 }
348
349 } // namespace zagryadskov_m_complex_spmm_ccs
350