-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathdriver.cpp
221 lines (200 loc) · 11 KB
/
driver.cpp
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
#include "globalIndexCreation.cpp"
// document id's on different nodes should have different names
int main(int argc, char *argv[])
{
MPI_Init(NULL,NULL);
double startLocalIndexTime, endLocalIndexTime;
startLocalIndexTime = MPI_Wtime();
int num_processes;
MPI_Comm_size(MPI_COMM_WORLD, &num_processes);
int my_rank;
MPI_Status status;
MPI_Comm_rank(MPI_COMM_WORLD,&my_rank);
map<string, bool> stopWordsIndex;
map<string,vector<docFreq> > localIndex;
ifstream fp("stopwords.txt");
string stopWord;
// create stop words map
while(getline(fp,stopWord))
{
stopWordsIndex[stopWord] = true;
}
fp.close();
// map<string,bool>::iterator stop_itr;
// for(stop_itr = stopWordsIndex.begin(); stop_itr != stopWordsIndex.end(); stop_itr++)
// {
// cout << stop_itr->first << "\n" << stop_itr->second << "\n";
// }
// create local index
char* dirPath = argv[1];
localIndex = createLocalIndex(dirPath,stopWordsIndex,my_rank);
// printLocalIndex(localIndex);
endLocalIndexTime = MPI_Wtime();
printf("Time taken for local Indexing by process %d: %lf seconds\n", my_rank, endLocalIndexTime - startLocalIndexTime);
// write local index to file
map<string, vector<docFreq> >::iterator map_itr;
string file_name = "localIndex" + to_string(my_rank) + ".txt";
ofstream localIndexFile(file_name,ios::out);
for(map_itr = localIndex.begin(); map_itr != localIndex.end(); map_itr++)
{
localIndexFile << (map_itr->first + ": ");
vector<docFreq>::iterator vec_itr;
for(vec_itr = (map_itr->second).begin(); vec_itr != (map_itr->second).end(); vec_itr++)
{
localIndexFile << ("("+to_string(vec_itr->docID)+", "+to_string(vec_itr->frequency)+") ");
}
localIndexFile << "\n";
}
// do parallel processing only if number of processes > 1 and number of processes is a power of 2
if(num_processes == 1)
{
MPI_Finalize();
return 0;
}
else if(ceil(log2(num_processes)) != floor(log2(num_processes)))
{
printf("Give number of processes as power of 2 for better scalability\n");
MPI_Finalize();
return 0;
}
// synchronization point
MPI_Barrier(MPI_COMM_WORLD);
// global merging
double startGlobalIndexTime, endGlobalIndexTime;
startGlobalIndexTime = MPI_Wtime();
int height_tree = (int)log2(num_processes);
int level = 1;
map<string,vector<docFreq> > mergedHashMapHalf = localIndex;
while(level <= height_tree)
{
int pow_2_level = (int)pow(2,level);
int pow_2_level_minus = (int)pow(2,level-1);
if(my_rank % pow_2_level_minus == 0)
{
serializedMap s_map_send;
// serialize the index in a process created so far
s_map_send = serializeLocalMap(mergedHashMapHalf);
serializedMap s_map_recv;
vector<int> sizes;
sizes.clear();
// send sizes of vectors so that on receiver side, memory is allocated for the vectors(buffers)
sizes.push_back(s_map_send.charsInWords.size());
sizes.push_back(s_map_send.indexOfStrings.size());
sizes.push_back(s_map_send.docIDArr.size());
vector<int> sizes_recv(3);
// half of the processes will send first then receive
if(my_rank % pow_2_level == 0)
{
MPI_Send(&sizes[0],sizes.size(),MPI_INT,(my_rank + pow_2_level_minus)%num_processes,0,MPI_COMM_WORLD);
MPI_Send(&s_map_send.charsInWords[0],s_map_send.charsInWords.size(),MPI_BYTE,(my_rank+pow_2_level_minus)%num_processes,1,MPI_COMM_WORLD);
MPI_Send(&s_map_send.indexOfStrings[0],s_map_send.indexOfStrings.size(),MPI_INT,(my_rank+pow_2_level_minus)%num_processes,2,MPI_COMM_WORLD);
MPI_Send(&s_map_send.docIDArr[0],s_map_send.docIDArr.size(),MPI_INT,(my_rank+pow_2_level_minus)%num_processes,3,MPI_COMM_WORLD);
MPI_Send(&s_map_send.FreqArr[0],s_map_send.FreqArr.size(),MPI_INT,(my_rank+pow_2_level_minus)%num_processes,4,MPI_COMM_WORLD);
MPI_Send(&s_map_send.indexDocFreq[0],s_map_send.indexDocFreq.size(),MPI_INT,(my_rank+pow_2_level_minus)%num_processes,5,MPI_COMM_WORLD);
MPI_Recv(&sizes_recv[0],3,MPI_INT,(my_rank+pow_2_level_minus)%num_processes,0,MPI_COMM_WORLD,&status);
s_map_recv.charsInWords.resize(sizes_recv[0]);
s_map_recv.indexOfStrings.resize(sizes_recv[1]);
s_map_recv.indexDocFreq.resize(sizes_recv[1]);
s_map_recv.docIDArr.resize(sizes_recv[2]);
s_map_recv.FreqArr.resize(sizes_recv[2]);
MPI_Recv(&s_map_recv.charsInWords[0], s_map_recv.charsInWords.size(), MPI_BYTE, (my_rank+pow_2_level_minus)%num_processes, 1, MPI_COMM_WORLD, &status);
MPI_Recv(&s_map_recv.indexOfStrings[0], s_map_recv.indexOfStrings.size(), MPI_INT, (my_rank+pow_2_level_minus)%num_processes, 2, MPI_COMM_WORLD, &status);
MPI_Recv(&s_map_recv.docIDArr[0], s_map_recv.docIDArr.size(), MPI_INT, (my_rank+pow_2_level_minus)%num_processes, 3, MPI_COMM_WORLD, &status);
MPI_Recv(&s_map_recv.FreqArr[0], s_map_recv.FreqArr.size(), MPI_INT, (my_rank+pow_2_level_minus)%num_processes, 4, MPI_COMM_WORLD, &status);
MPI_Recv(&s_map_recv.indexDocFreq[0], s_map_recv.indexDocFreq.size(), MPI_INT, (my_rank+pow_2_level_minus)%num_processes, 5, MPI_COMM_WORLD, &status);
}
// half of the processes will receive first then send
else
{
MPI_Recv(&sizes_recv[0],3,MPI_INT,(my_rank-pow_2_level_minus)%num_processes,0,MPI_COMM_WORLD,&status);
s_map_recv.charsInWords.resize(sizes_recv[0]);
s_map_recv.indexOfStrings.resize(sizes_recv[1]);
s_map_recv.indexDocFreq.resize(sizes_recv[1]);
s_map_recv.docIDArr.resize(sizes_recv[2]);
s_map_recv.FreqArr.resize(sizes_recv[2]);
MPI_Recv(&s_map_recv.charsInWords[0], s_map_recv.charsInWords.size(), MPI_BYTE, (my_rank-pow_2_level_minus)%num_processes, 1, MPI_COMM_WORLD, &status);
MPI_Recv(&s_map_recv.indexOfStrings[0], s_map_recv.indexOfStrings.size(), MPI_INT, (my_rank-pow_2_level_minus)%num_processes, 2, MPI_COMM_WORLD, &status);
MPI_Recv(&s_map_recv.docIDArr[0], s_map_recv.docIDArr.size(), MPI_INT, (my_rank-pow_2_level_minus)%num_processes, 3, MPI_COMM_WORLD, &status);
MPI_Recv(&s_map_recv.FreqArr[0], s_map_recv.FreqArr.size(), MPI_INT, (my_rank-pow_2_level_minus)%num_processes, 4, MPI_COMM_WORLD, &status);
MPI_Recv(&s_map_recv.indexDocFreq[0], s_map_recv.indexDocFreq.size(), MPI_INT, (my_rank-pow_2_level_minus)%num_processes, 5, MPI_COMM_WORLD, &status);
MPI_Send(&sizes[0],sizes.size(),MPI_INT,(my_rank-pow_2_level_minus)%num_processes,0,MPI_COMM_WORLD);
MPI_Send(&s_map_send.charsInWords[0],s_map_send.charsInWords.size(),MPI_BYTE,(my_rank-pow_2_level_minus)%num_processes,1,MPI_COMM_WORLD);
MPI_Send(&s_map_send.indexOfStrings[0],s_map_send.indexOfStrings.size(),MPI_INT,(my_rank-pow_2_level_minus)%num_processes,2,MPI_COMM_WORLD);
MPI_Send(&s_map_send.docIDArr[0],s_map_send.docIDArr.size(),MPI_INT,(my_rank-pow_2_level_minus)%num_processes,3,MPI_COMM_WORLD);
MPI_Send(&s_map_send.FreqArr[0],s_map_send.FreqArr.size(),MPI_INT,(my_rank-pow_2_level_minus)%num_processes,4,MPI_COMM_WORLD);
MPI_Send(&s_map_send.indexDocFreq[0],s_map_send.indexDocFreq.size(),MPI_INT,(my_rank-pow_2_level_minus)%num_processes,5,MPI_COMM_WORLD);
}
// deserialize the map received
map<string, pair<int,int> > de_map_recv = deserializeMap(s_map_recv);
// map<string, pair<int,int> >::iterator de_itr;
// for(de_itr = de_map_recv.begin(); de_itr != de_map_recv.end(); de_itr++)
// {
// printf("%s :(%d,%d) process:%d\n",(de_itr->first).c_str(),(de_itr->second).first,(de_itr->second).second,my_rank);
// }
// merge the local map and the map received from neighbour process create half a final map
mergedHashMapHalf = mergeHashMaps(mergedHashMapHalf,de_map_recv,s_map_recv.docIDArr,s_map_recv.FreqArr,my_rank,level,num_processes); // other half will be sent by neighbour process
// map<string,vector<docFreq> >::iterator mh_itr;
// for(mh_itr = mergedHashMapHalf.begin(); mh_itr != mergedHashMapHalf.end(); mh_itr++)
// {
// // printf("%s(%d) :\n",mh_itr->first,my_rank);
// vector<docFreq>::iterator v_itr;
// for(v_itr = (mh_itr->second).begin(); v_itr != (mh_itr->second).end();v_itr++)
// {
// printf("%s(%d)(%d) process:%d\n",(mh_itr->first).c_str(),v_itr->docID,v_itr->frequency,my_rank);
// }
// }
// one process will send the map after serializing it
if(my_rank%pow_2_level != 0)
{
s_map_send = serializeLocalMap(mergedHashMapHalf);
sizes.clear();
sizes.push_back(s_map_send.charsInWords.size());
sizes.push_back(s_map_send.indexOfStrings.size());
sizes.push_back(s_map_send.docIDArr.size());
MPI_Send(&sizes[0],sizes.size(),MPI_INT,(my_rank-pow_2_level_minus)%num_processes,0,MPI_COMM_WORLD);
MPI_Send(&s_map_send.charsInWords[0],s_map_send.charsInWords.size(),MPI_BYTE,(my_rank-pow_2_level_minus)%num_processes,1,MPI_COMM_WORLD);
MPI_Send(&s_map_send.indexOfStrings[0],s_map_send.indexOfStrings.size(),MPI_INT,(my_rank-pow_2_level_minus)%num_processes,2,MPI_COMM_WORLD);
MPI_Send(&s_map_send.docIDArr[0],s_map_send.docIDArr.size(),MPI_INT,(my_rank-pow_2_level_minus)%num_processes,3,MPI_COMM_WORLD);
MPI_Send(&s_map_send.FreqArr[0],s_map_send.FreqArr.size(),MPI_INT,(my_rank-pow_2_level_minus)%num_processes,4,MPI_COMM_WORLD);
MPI_Send(&s_map_send.indexDocFreq[0],s_map_send.indexDocFreq.size(),MPI_INT,(my_rank-pow_2_level_minus)%num_processes,5,MPI_COMM_WORLD);
}
// other process will receive the map and do combining of the 2 half maps
else
{
MPI_Recv(&sizes_recv[0],3,MPI_INT,(my_rank+pow_2_level_minus)%num_processes,0,MPI_COMM_WORLD,&status);
s_map_recv.charsInWords.resize(sizes_recv[0]);
s_map_recv.indexOfStrings.resize(sizes_recv[1]);
s_map_recv.indexDocFreq.resize(sizes_recv[1]);
s_map_recv.docIDArr.resize(sizes_recv[2]);
s_map_recv.FreqArr.resize(sizes_recv[2]);
MPI_Recv(&s_map_recv.charsInWords[0], s_map_recv.charsInWords.size(), MPI_BYTE, (my_rank+pow_2_level_minus)%num_processes, 1, MPI_COMM_WORLD, &status);
MPI_Recv(&s_map_recv.indexOfStrings[0], s_map_recv.indexOfStrings.size(), MPI_INT, (my_rank+pow_2_level_minus)%num_processes, 2, MPI_COMM_WORLD, &status);
MPI_Recv(&s_map_recv.docIDArr[0], s_map_recv.docIDArr.size(), MPI_INT, (my_rank+pow_2_level_minus)%num_processes, 3, MPI_COMM_WORLD, &status);
MPI_Recv(&s_map_recv.FreqArr[0], s_map_recv.FreqArr.size(), MPI_INT, (my_rank+pow_2_level_minus)%num_processes, 4, MPI_COMM_WORLD, &status);
MPI_Recv(&s_map_recv.indexDocFreq[0], s_map_recv.indexDocFreq.size(), MPI_INT, (my_rank+pow_2_level_minus)%num_processes, 5, MPI_COMM_WORLD, &status);
mergedHashMapHalf = deserializeAndMerge(mergedHashMapHalf,s_map_recv);
}
}
level++;
}
endGlobalIndexTime = MPI_Wtime();
// write global index in a file
if(my_rank == 0)
{
printf("\nTime taken for creating global index %lf seconds\n", endGlobalIndexTime - startGlobalIndexTime);
file_name = "globalIndex.txt";
ofstream globalIndexFile(file_name,ios::out);
for(map_itr = mergedHashMapHalf.begin(); map_itr != mergedHashMapHalf.end(); map_itr++)
{
globalIndexFile << (map_itr->first + ": ");
vector<docFreq>::iterator vec_itr;
for(vec_itr = (map_itr->second).begin(); vec_itr != (map_itr->second).end(); vec_itr++)
{
globalIndexFile << ("("+to_string(vec_itr->docID)+", "+to_string(vec_itr->frequency)+") ");
}
globalIndexFile << "\n";
}
}
MPI_Finalize();
return 0;
}