#include <algorithm> #include <climits> #include <cstdio> #include <cstdlib> #include <string> #include <vector> #include "message.h" #include "poszukiwania.h" using namespace std; long long S, M; struct solved_partition { solved_partition(long long start_, long long end_, long long h_) : start(start_), end(end_), h(h_) { } bool operator < (const solved_partition& x) const { if (start == x.start) { if (end == x.end) { return h < x.h; } return end < x.end; } return start < x.start; } long long start, end, h; }; #define H1 3425706325711LL long long fast_pow(long long x, long long n) { long long v = 1LL; while (n > 0) { if (n & 1) { v = v * x; } x = x * x; n /= 2LL; } return v; } long long HashS(long long start, long long end, long long p) { long long h = 0; for (long long i = start; i <= end; i++) { h = (h * p) + SignalAt(i); } return h; } long long HashM(long long start, long long end, long long p) { long long h = 0; for (long long i = start; i <= end; i++) { h = (h * p) + SeqAt(i); } return h; } int DoTheCounting(long long start, long long end, long long h_s, long long h_m, long long p) { // Calculate p^S, required during the recalculation of the sequence // substring hash. long long p_pow_S = fast_pow(p, S); // Iterate over all requested offsets. int sum = 0; for (long long i = start; i <= end && i <= M - S + 1; i++) { if (h_m == h_s) { sum++; } // Adjust the hash to move the substring by one to the right. h_m *= p; h_m += SeqAt(i + S); h_m -= SeqAt(i) * p_pow_S; } return sum; } long long CalculatePartialHashM(const vector<solved_partition>& parts, size_t idx) { long long count = 0; long long h = 0; long long local_start = parts[idx].start; while (count < S && idx < parts.size()) { long long expected_power = S - (parts[idx].start - local_start + 1); long long actual_power = (parts[idx].end - parts[idx].start + 1) - 1; h += parts[idx].h * fast_pow(H1, expected_power - actual_power); count += (parts[idx].end - parts[idx].start + 1); idx++; } return h; } int main() { // Get the S, M sequence lengths. S = SignalLength(); M = SeqLength(); if (M <= 1000000) { // If M (and therefore S) is reasonably small, just perform the calculations // locally in task 0, by calculating maximum pref-suff for S#M. if (MyNodeId() == 0) { vector<long long> w(1 + S + 1 + M); vector<long long> t(S + 1 + M); w[0] = -1; for (long long i = 0; i < S; i++) { t[i] = SignalAt(i + 1); } t[S] = INT_MAX; for (long long i = 0; i < M; i++) { t[S + 1 + i] = SeqAt(i + 1); } for (long long i = 0, j = -1; i < S + 1 + M; i++, j++, w[i] = j) { while ((j >= 0) && (t[i] != t[j])) { j = w[j]; } } int sum = 0; for (long long i = 0; i < S + 1 + M; i++) { if (w[i + 1] == S) { sum++; } } printf("%d\n", sum); } else { // Do nothing in other tasks, just exit. } } else { long long items_per_task = M / (NumberOfNodes() - 1); if (S <= items_per_task) { // S is smaller than the workload each task has to handle in M. Therefore, // just assign the direct work to the machines, wait for the results, // aggregate them and print out. if (MyNodeId() == 0) { int sum = 0; int tasks_done = 0; while (tasks_done < NumberOfNodes() - 1) { int source = Receive(-1); sum += GetInt(source); tasks_done++; } printf("%d\n", sum); } else { // This is where actual work is done. Calculate the range of offsets // which needs to be covered by this task. long long start = 1 + (M * (MyNodeId() - 1)) / (NumberOfNodes() - 1); long long end; if (MyNodeId() != NumberOfNodes() - 1) { end = ((M * MyNodeId()) / (NumberOfNodes() - 1)); } else { end = M - S; } // Calculate the overall hash for S. long long h_s = HashS(1, S, H1); // Calculate an initial hash for M. long long h_m = HashM(start, start + S - 1, H1); PutInt(0, DoTheCounting(start, end, h_s, h_m, H1)); Send(0); } } else { // S is larger than a statistical workload of M for a single machine. // Therefore, we also have to split the hash calculation of S between // instances. if (MyNodeId() == 0) { // Partition S and assign tasks to compute hashes of substrings. for (int i = 1; i < NumberOfNodes(); i++) { long long start = 1 + (S * (i - 1)) / (NumberOfNodes() - 1); long long end; if (MyNodeId() != NumberOfNodes() - 1) { end = ((S * i) / (NumberOfNodes() - 1)); } else { end = S; } PutInt(i, 1); // 1: there's work incoming! PutInt(i, 0); // 0: we're interested in a S hash. PutLL(i, start); // starting index to hash. PutLL(i, end); // end index to hash. Send(i); } // Receive the results and reassemble the hash. long long h_s = 0; int tasks_done = 0; while (tasks_done < NumberOfNodes() - 1) { int source = Receive(-1); long long start = GetLL(source); long long end = GetLL(source); long long h = GetLL(source); long long expected_power = S - start; long long actual_power = (end - start + 1) - 1; //printf("Received hash: [%lld, %lld]: %lld\n", start, end, h); //printf("Expected power: %lld, actual power: %lld\n", expected_power, actual_power); h_s += h * fast_pow(H1, expected_power - actual_power); tasks_done++; } //printf("Reassembled S hash: %lld\n", h_s); // Partition M and assign tasks to compute hashes of substrings. vector<long long> partitions; long long count = 0; long long S_left = S; while (count < M) { long long part = min(min(items_per_task, S_left), M - count); partitions.push_back(part); //printf("part: %lld\n", part); S_left -= part; if (S_left == 0) { S_left = S; } count += part; } //printf("Calculation tasks: %d, partitions: %lu\n", NumberOfNodes() - 1, partitions.size()); // Assign the partitions to tasks for calculation. long long sum = 0; for (size_t i = 0; i < partitions.size(); i++) { int task_id = 1 + ((i * (NumberOfNodes() - 1)) / partitions.size()); //printf("Assigning partition %lu to task %d\n", i, task_id); PutInt(task_id, 1); // 1: there's work incoming! PutInt(task_id, 1); // 1: we're interested in a M hash. PutLL(task_id, sum + 1); // starting index to hash. PutLL(task_id, sum + partitions[i]); // end index to hash. Send(task_id); sum += partitions[i]; } // Let all tasks know that it's the end of work. for (int i = 1; i < NumberOfNodes(); i++) { PutInt(i, 0); // 0: no more hashing work, proceed to the next phase. Send(i); } // Receive all partial M hashes. size_t partitions_done = 0; vector<solved_partition> solved_partitions; while (partitions_done < partitions.size()) { int source = Receive(-1); long long start = GetLL(source); long long end = GetLL(source); long long h = GetLL(source); solved_partitions.emplace_back(start, end, h); partitions_done++; } // Sort the partitions. sort(solved_partitions.begin(), solved_partitions.end()); //for (size_t i = 0; i < solved_partitions.size(); i++) { // printf("Solved partition [%lld, %lld]: %lld\n", // solved_partitions[i].start, solved_partitions[i].end, solved_partitions[i].h); //} // Count the partitions which actually need to be assigned. size_t assignable_partitions = 0; for (size_t i = 0; i < solved_partitions.size(); i++) { if (solved_partitions[i].start <= M - S + 1) { assignable_partitions++; } else { break; } } //printf("Assignable partitions: %d\n", assignable_partitions); // Assign partitions to tasks. for (size_t i = 0; i < assignable_partitions; i++) { int task_id = 1 + ((i * (NumberOfNodes() - 1)) / assignable_partitions); long long h_m = CalculatePartialHashM(solved_partitions, i); //printf("Assigning partition [%lld, %lld] to task %d for counting with hash %lld.\n", // solved_partitions[i].start, solved_partitions[i].end, task_id, h_m); PutInt(task_id, 1); // 1: there's work incoming! PutLL(task_id, solved_partitions[i].start); // starting index for counting. PutLL(task_id, solved_partitions[i].end); // ending index for couting. PutLL(task_id, h_s); // overall S hash. PutLL(task_id, h_m); // initial M hash. Send(task_id); } // Let all tasks know that now it's really the end of work. for (int i = 1; i < NumberOfNodes(); i++) { PutInt(i, 0); // no more couting work. Send(i); } // Count results for all requests. int result = 0; partitions_done = 0; while (partitions_done < assignable_partitions) { int source = Receive(-1); result += GetInt(source); partitions_done++; } // Print out the result. printf("%d\n", result); } else { // Receive hashing requests from the master. int more_work = 0; do { Receive(0); more_work = GetInt(0); if (more_work) { int type = GetInt(0); long long start = GetLL(0); long long end = GetLL(0); //printf("[%d] Received [%d, %lld, %lld] request from master.\n", // MyNodeId(), type, start, end); PutLL(0, start); PutLL(0, end); if (type == 0) { PutLL(0, HashS(start, end, H1)); } else { PutLL(0, HashM(start, end, H1)); } Send(0); } } while (more_work != 0); // Receive counting requests from the master. more_work = 0; do { Receive(0); more_work = GetInt(0); if (more_work) { long long start = GetLL(0); long long end = GetLL(0); long long h_s = GetLL(0); long long h_m = GetLL(0); //printf("[%d] Received [%lld, %lld, %lld, %lld] request from master.\n", // MyNodeId(), start, end, h_s, h_m); PutInt(0, DoTheCounting(start, end, h_s, h_m, H1)); Send(0); } } while (more_work != 0); } } } return 0; }
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 | #include <algorithm> #include <climits> #include <cstdio> #include <cstdlib> #include <string> #include <vector> #include "message.h" #include "poszukiwania.h" using namespace std; long long S, M; struct solved_partition { solved_partition(long long start_, long long end_, long long h_) : start(start_), end(end_), h(h_) { } bool operator < (const solved_partition& x) const { if (start == x.start) { if (end == x.end) { return h < x.h; } return end < x.end; } return start < x.start; } long long start, end, h; }; #define H1 3425706325711LL long long fast_pow(long long x, long long n) { long long v = 1LL; while (n > 0) { if (n & 1) { v = v * x; } x = x * x; n /= 2LL; } return v; } long long HashS(long long start, long long end, long long p) { long long h = 0; for (long long i = start; i <= end; i++) { h = (h * p) + SignalAt(i); } return h; } long long HashM(long long start, long long end, long long p) { long long h = 0; for (long long i = start; i <= end; i++) { h = (h * p) + SeqAt(i); } return h; } int DoTheCounting(long long start, long long end, long long h_s, long long h_m, long long p) { // Calculate p^S, required during the recalculation of the sequence // substring hash. long long p_pow_S = fast_pow(p, S); // Iterate over all requested offsets. int sum = 0; for (long long i = start; i <= end && i <= M - S + 1; i++) { if (h_m == h_s) { sum++; } // Adjust the hash to move the substring by one to the right. h_m *= p; h_m += SeqAt(i + S); h_m -= SeqAt(i) * p_pow_S; } return sum; } long long CalculatePartialHashM(const vector<solved_partition>& parts, size_t idx) { long long count = 0; long long h = 0; long long local_start = parts[idx].start; while (count < S && idx < parts.size()) { long long expected_power = S - (parts[idx].start - local_start + 1); long long actual_power = (parts[idx].end - parts[idx].start + 1) - 1; h += parts[idx].h * fast_pow(H1, expected_power - actual_power); count += (parts[idx].end - parts[idx].start + 1); idx++; } return h; } int main() { // Get the S, M sequence lengths. S = SignalLength(); M = SeqLength(); if (M <= 1000000) { // If M (and therefore S) is reasonably small, just perform the calculations // locally in task 0, by calculating maximum pref-suff for S#M. if (MyNodeId() == 0) { vector<long long> w(1 + S + 1 + M); vector<long long> t(S + 1 + M); w[0] = -1; for (long long i = 0; i < S; i++) { t[i] = SignalAt(i + 1); } t[S] = INT_MAX; for (long long i = 0; i < M; i++) { t[S + 1 + i] = SeqAt(i + 1); } for (long long i = 0, j = -1; i < S + 1 + M; i++, j++, w[i] = j) { while ((j >= 0) && (t[i] != t[j])) { j = w[j]; } } int sum = 0; for (long long i = 0; i < S + 1 + M; i++) { if (w[i + 1] == S) { sum++; } } printf("%d\n", sum); } else { // Do nothing in other tasks, just exit. } } else { long long items_per_task = M / (NumberOfNodes() - 1); if (S <= items_per_task) { // S is smaller than the workload each task has to handle in M. Therefore, // just assign the direct work to the machines, wait for the results, // aggregate them and print out. if (MyNodeId() == 0) { int sum = 0; int tasks_done = 0; while (tasks_done < NumberOfNodes() - 1) { int source = Receive(-1); sum += GetInt(source); tasks_done++; } printf("%d\n", sum); } else { // This is where actual work is done. Calculate the range of offsets // which needs to be covered by this task. long long start = 1 + (M * (MyNodeId() - 1)) / (NumberOfNodes() - 1); long long end; if (MyNodeId() != NumberOfNodes() - 1) { end = ((M * MyNodeId()) / (NumberOfNodes() - 1)); } else { end = M - S; } // Calculate the overall hash for S. long long h_s = HashS(1, S, H1); // Calculate an initial hash for M. long long h_m = HashM(start, start + S - 1, H1); PutInt(0, DoTheCounting(start, end, h_s, h_m, H1)); Send(0); } } else { // S is larger than a statistical workload of M for a single machine. // Therefore, we also have to split the hash calculation of S between // instances. if (MyNodeId() == 0) { // Partition S and assign tasks to compute hashes of substrings. for (int i = 1; i < NumberOfNodes(); i++) { long long start = 1 + (S * (i - 1)) / (NumberOfNodes() - 1); long long end; if (MyNodeId() != NumberOfNodes() - 1) { end = ((S * i) / (NumberOfNodes() - 1)); } else { end = S; } PutInt(i, 1); // 1: there's work incoming! PutInt(i, 0); // 0: we're interested in a S hash. PutLL(i, start); // starting index to hash. PutLL(i, end); // end index to hash. Send(i); } // Receive the results and reassemble the hash. long long h_s = 0; int tasks_done = 0; while (tasks_done < NumberOfNodes() - 1) { int source = Receive(-1); long long start = GetLL(source); long long end = GetLL(source); long long h = GetLL(source); long long expected_power = S - start; long long actual_power = (end - start + 1) - 1; //printf("Received hash: [%lld, %lld]: %lld\n", start, end, h); //printf("Expected power: %lld, actual power: %lld\n", expected_power, actual_power); h_s += h * fast_pow(H1, expected_power - actual_power); tasks_done++; } //printf("Reassembled S hash: %lld\n", h_s); // Partition M and assign tasks to compute hashes of substrings. vector<long long> partitions; long long count = 0; long long S_left = S; while (count < M) { long long part = min(min(items_per_task, S_left), M - count); partitions.push_back(part); //printf("part: %lld\n", part); S_left -= part; if (S_left == 0) { S_left = S; } count += part; } //printf("Calculation tasks: %d, partitions: %lu\n", NumberOfNodes() - 1, partitions.size()); // Assign the partitions to tasks for calculation. long long sum = 0; for (size_t i = 0; i < partitions.size(); i++) { int task_id = 1 + ((i * (NumberOfNodes() - 1)) / partitions.size()); //printf("Assigning partition %lu to task %d\n", i, task_id); PutInt(task_id, 1); // 1: there's work incoming! PutInt(task_id, 1); // 1: we're interested in a M hash. PutLL(task_id, sum + 1); // starting index to hash. PutLL(task_id, sum + partitions[i]); // end index to hash. Send(task_id); sum += partitions[i]; } // Let all tasks know that it's the end of work. for (int i = 1; i < NumberOfNodes(); i++) { PutInt(i, 0); // 0: no more hashing work, proceed to the next phase. Send(i); } // Receive all partial M hashes. size_t partitions_done = 0; vector<solved_partition> solved_partitions; while (partitions_done < partitions.size()) { int source = Receive(-1); long long start = GetLL(source); long long end = GetLL(source); long long h = GetLL(source); solved_partitions.emplace_back(start, end, h); partitions_done++; } // Sort the partitions. sort(solved_partitions.begin(), solved_partitions.end()); //for (size_t i = 0; i < solved_partitions.size(); i++) { // printf("Solved partition [%lld, %lld]: %lld\n", // solved_partitions[i].start, solved_partitions[i].end, solved_partitions[i].h); //} // Count the partitions which actually need to be assigned. size_t assignable_partitions = 0; for (size_t i = 0; i < solved_partitions.size(); i++) { if (solved_partitions[i].start <= M - S + 1) { assignable_partitions++; } else { break; } } //printf("Assignable partitions: %d\n", assignable_partitions); // Assign partitions to tasks. for (size_t i = 0; i < assignable_partitions; i++) { int task_id = 1 + ((i * (NumberOfNodes() - 1)) / assignable_partitions); long long h_m = CalculatePartialHashM(solved_partitions, i); //printf("Assigning partition [%lld, %lld] to task %d for counting with hash %lld.\n", // solved_partitions[i].start, solved_partitions[i].end, task_id, h_m); PutInt(task_id, 1); // 1: there's work incoming! PutLL(task_id, solved_partitions[i].start); // starting index for counting. PutLL(task_id, solved_partitions[i].end); // ending index for couting. PutLL(task_id, h_s); // overall S hash. PutLL(task_id, h_m); // initial M hash. Send(task_id); } // Let all tasks know that now it's really the end of work. for (int i = 1; i < NumberOfNodes(); i++) { PutInt(i, 0); // no more couting work. Send(i); } // Count results for all requests. int result = 0; partitions_done = 0; while (partitions_done < assignable_partitions) { int source = Receive(-1); result += GetInt(source); partitions_done++; } // Print out the result. printf("%d\n", result); } else { // Receive hashing requests from the master. int more_work = 0; do { Receive(0); more_work = GetInt(0); if (more_work) { int type = GetInt(0); long long start = GetLL(0); long long end = GetLL(0); //printf("[%d] Received [%d, %lld, %lld] request from master.\n", // MyNodeId(), type, start, end); PutLL(0, start); PutLL(0, end); if (type == 0) { PutLL(0, HashS(start, end, H1)); } else { PutLL(0, HashM(start, end, H1)); } Send(0); } } while (more_work != 0); // Receive counting requests from the master. more_work = 0; do { Receive(0); more_work = GetInt(0); if (more_work) { long long start = GetLL(0); long long end = GetLL(0); long long h_s = GetLL(0); long long h_m = GetLL(0); //printf("[%d] Received [%lld, %lld, %lld, %lld] request from master.\n", // MyNodeId(), start, end, h_s, h_m); PutInt(0, DoTheCounting(start, end, h_s, h_m, H1)); Send(0); } } while (more_work != 0); } } } return 0; } |