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
|
// Copyright (c) 2020 Robert Vaser
// Combination of ThreadPool implementation by progschj and
// task stealing by Sean Parent
#ifndef THREAD_POOL_THREAD_POOL_HPP_
#define THREAD_POOL_THREAD_POOL_HPP_
#include <algorithm>
#include <atomic>
#include <functional>
#include <future> // NOLINT
#include <memory>
#include <queue>
#include <thread> // NOLINT
#include <unordered_map>
#include <utility>
#include <vector>
namespace thread_pool {
class ThreadPool {
public:
explicit ThreadPool(
std::size_t num_threads = std::thread::hardware_concurrency())
: threads_(),
thread_map_(),
queues_(std::max(size_t(1), num_threads)),
task_id_(0) {
for (std::size_t i = 0; i != queues_.size(); ++i) {
threads_.emplace_back([this, i] () -> void { Task(i); });
thread_map_.emplace(threads_.back().get_id(), i);
}
}
ThreadPool(const ThreadPool&) = delete;
ThreadPool& operator=(const ThreadPool&) = delete;
ThreadPool(ThreadPool&&) = delete;
ThreadPool& operator=(ThreadPool&&) = delete;
~ThreadPool() {
for (auto& it : queues_) {
it.Done();
}
for (auto& it : threads_) {
it.join();
}
}
std::size_t num_threads() const {
return threads_.size();
}
const std::unordered_map<std::thread::id, std::size_t>& thread_map() const {
return thread_map_;
}
template<typename T, typename... Ts>
auto Submit(T&& routine, Ts&&... params)
-> std::future<typename std::result_of<T(Ts...)>::type> {
auto task = std::make_shared<std::packaged_task<typename std::result_of<T(Ts...)>::type()>>( // NOLINT
std::bind(std::forward<T>(routine), std::forward<Ts>(params)...));
auto task_result = task->get_future();
auto task_wrapper = [task] () {
(*task)();
};
auto task_id = task_id_++;
bool is_submitted = false;
for (std::size_t i = 0; i != queues_.size() * 42; ++i) {
if (queues_[(task_id + i) % queues_.size()].TryPush(task_wrapper)) {
is_submitted = true;
break;
}
}
if (!is_submitted) {
queues_[task_id % queues_.size()].Push(task_wrapper);
}
return task_result;
}
private:
void Task(std::size_t thread_id) {
while (true) {
std::function<void()> task;
for (std::size_t i = 0; i != queues_.size(); ++i) {
if (queues_[(thread_id + i) % queues_.size()].TryPop(&task)) {
break;
}
}
if (!task && !queues_[thread_id].Pop(&task)) {
break;
}
task();
}
}
struct TaskQueue {
public:
template<typename F>
void Push(F&& f) {
{
std::unique_lock<std::mutex> lock(mutex);
queue.emplace(std::forward<F>(f));
}
is_ready.notify_one();
}
bool Pop(std::function<void()>* f) {
std::unique_lock<std::mutex> lock(mutex);
while (queue.empty() && !is_done) {
is_ready.wait(lock);
}
if (queue.empty()) {
return false;
}
*f = std::move(queue.front());
queue.pop();
return true;
}
template<typename F>
bool TryPush(F&& f) {
{
std::unique_lock<std::mutex> lock(mutex, std::try_to_lock);
if (!lock) {
return false;
}
queue.emplace(std::forward<F>(f));
}
is_ready.notify_one();
return true;
}
bool TryPop(std::function<void()>* f) {
std::unique_lock<std::mutex> lock(mutex, std::try_to_lock);
if (!lock || queue.empty()) {
return false;
}
*f = std::move(queue.front());
queue.pop();
return true;
}
void Done() {
{
std::unique_lock<std::mutex> lock(mutex);
is_done = true;
}
is_ready.notify_all();
}
std::queue<std::function<void()>> queue;
std::mutex mutex;
std::condition_variable is_ready;
bool is_done = false;
};
std::vector<std::thread> threads_;
std::unordered_map<std::thread::id, std::size_t> thread_map_;
std::vector<TaskQueue> queues_;
std::atomic<std::size_t> task_id_;
};
} // namespace thread_pool
#endif // THREAD_POOL_THREAD_POOL_HPP_
|
