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
|
// example graciously provided @samuelpmish
// ----------------------------------------
//
// Getting rid of the mutexes for read access
//
// This example demonstrated how to populate a parallel_flat_hash_map from multiple
// concurrent threads (The map is protected by internal mutexes), but then doing a
// swap to get rid of the mutexes (and all locking) for accessing the same hash_map
// in `read` only mode, again concurrently from multiple threads.
// --------------------------------------------------------------------------------
#include <random>
#include <iostream>
#include <unordered_map>
#include <unordered_set>
#include "parallel_hashmap/phmap.h"
///////////////////////////////////////////////////////////////////////////////
#include <chrono>
class timer {
typedef std::chrono::high_resolution_clock::time_point time_point;
typedef std::chrono::duration<double> duration_type;
public:
void start() { then = std::chrono::high_resolution_clock::now(); }
void stop() { now = std::chrono::high_resolution_clock::now(); }
double elapsed() { return std::chrono::duration_cast<duration_type>(now - then).count(); }
private:
time_point then, now;
};
///////////////////////////////////////////////////////////////////////////////
#include <thread>
struct threadpool {
std::vector< uint64_t > partition(uint64_t n) {
uint64_t quotient = n / num_threads;
uint64_t remainder = n % num_threads;
std::vector< uint64_t > blocks(num_threads + 1);
blocks[0] = 0;
for (int i = 1; i < num_threads + 1; i++) {
if (remainder > 0) {
blocks[i] = blocks[i-1] + quotient + 1;
remainder--;
} else {
blocks[i] = blocks[i-1] + quotient;
}
}
return blocks;
}
threadpool(int n) : num_threads(n) {}
template < typename lambda >
void parallel_for(uint64_t n, const lambda & f) {
std::vector< uint64_t > blocks = partition(n);
for (int tid = 0; tid < num_threads; tid++) {
threads.push_back(std::thread([&](uint64_t i0) {
for (uint64_t i = blocks[i0]; i < blocks[i0+1]; i++) {
f(i);
}
}, tid));
}
for (int i = 0; i < num_threads; i++) {
threads[i].join();
}
threads.clear();
}
int num_threads;
std::vector< std::thread > threads;
};
///////////////////////////////////////////////////////////////////////////////
template < int n >
using pmap = phmap::parallel_flat_hash_map<
uint64_t,
uint64_t,
std::hash<uint64_t>,
std::equal_to<uint64_t>,
std::allocator<std::pair<const uint64_t, uint64_t>>,
n,
std::mutex >;
template < int n >
using pmap_nullmutex = phmap::parallel_flat_hash_map<
uint64_t,
uint64_t,
std::hash<uint64_t>,
std::equal_to<uint64_t>,
std::allocator<std::pair<const uint64_t, uint64_t>>,
n,
phmap::NullMutex >;
template < typename Map, typename Map_nomutex >
void renumber(const std::vector< uint64_t > & vertex_ids,
std::vector< std::array< uint64_t, 4 > > elements,
int num_threads) {
bool supports_parallel_insertion =
!std::is_same< Map, std::unordered_map<uint64_t, uint64_t> >::value;
Map new_ids;
std::atomic< uint64_t > new_id{ 0 };
timer stopwatch;
threadpool pool((supports_parallel_insertion) ? num_threads : 1);
stopwatch.start();
new_ids.reserve(vertex_ids.size() * 110 / 100);
pool.parallel_for(vertex_ids.size(), [&](uint64_t i){
auto id = new_id++;
new_ids[vertex_ids[i]] = id;
});
stopwatch.stop();
std::cout << stopwatch.elapsed() * 1000 << "ms ";
pool.num_threads = num_threads;
stopwatch.start();
Map_nomutex new_ids_nc;
new_ids_nc.swap(new_ids);
pool.parallel_for(elements.size(), [&](uint64_t i) {
auto & elem = elements[i];
elem = { new_ids_nc.at(elem[0]),
new_ids_nc.at(elem[1]),
new_ids_nc.at(elem[2]),
new_ids_nc.at(elem[3]) };
});
stopwatch.stop();
std::cout << stopwatch.elapsed() * 1000 << "ms" << std::endl;
}
int main() {
uint64_t nvertices = 5000000;
uint64_t nelements = 25000000;
std::random_device rd; // a seed source for the random number engine
std::mt19937 gen(rd()); // mersenne_twister_engine seeded with rd()
std::uniform_int_distribution<uint64_t> vertex_id_dist(0, uint64_t(1) << 35);
std::uniform_int_distribution<uint64_t> elem_id_dist(0, nvertices-1);
std::cout << "generating dataset ." << std::flush;
std::vector< uint64_t > vertex_ids(nvertices);
for (uint64_t i = 0; i < nvertices; i++) {
vertex_ids[i] = vertex_id_dist(gen);
}
std::cout << "." << std::flush;
std::vector< std::array<uint64_t, 4> > elements(nelements);
for (uint64_t i = 0; i < nelements; i++) {
elements[i] = {
vertex_ids[elem_id_dist(gen)],
vertex_ids[elem_id_dist(gen)],
vertex_ids[elem_id_dist(gen)],
vertex_ids[elem_id_dist(gen)]
};
}
std::cout << " done" << std::endl;
using stdmap = std::unordered_map<uint64_t, uint64_t>;
std::cout << "std::unordered_map, 1 thread: ";
renumber< stdmap, stdmap >(vertex_ids, elements, 1);
std::cout << "std::unordered_map, 32 thread (single threaded insertion): ";
renumber< stdmap, stdmap >(vertex_ids, elements, 32);
std::cout << "pmap4, 1 thread: ";
renumber< pmap<4>, pmap_nullmutex<4> >(vertex_ids, elements, 1);
std::cout << "pmap4, 32 threads: ";
renumber< pmap<4>, pmap_nullmutex<4> >(vertex_ids, elements, 32);
std::cout << "pmap6, 1 thread: ";
renumber< pmap<6>, pmap_nullmutex<6> >(vertex_ids, elements, 1);
std::cout << "pmap6, 32 threads: ";
renumber< pmap<6>, pmap_nullmutex<6> >(vertex_ids, elements, 32);
}
|
