반응형

우선 Unsafe Region이 무엇인지 파악하기 위해 아래 링크를 참조해보자.


http://www.crocus.co.kr/562



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
#include <iostream>
#include <cstdio>
#include <queue>
#include <thread>
 
using namespace std;
 
int ans;
void func() {
    for (int i = 1; i <= 1000000; i++)
        ans++;
}
 
int main() {
    thread a(func);
    thread b(func);
    thread c(func);
 
    a.join();
    b.join();
    c.join();
    
    cout << "최종 ans :: " << ans << endl;
    return 0;
}
//                                                       This source code Copyright belongs to Crocus
//                                                        If you want to see more? click here >>
Crocus


이 코드를 보면 ans가 Unsafe Region이 된다.


Unsafe Region이 될 수 밖에 없는 이유는


1. 만약 a스레드가 ans = 1이고 ans++를 할 때 

2. 어셈블리 과정에서 ans + 1을 하려고 하는 순간 컨텍스트 스위칭을 당하게 되고 

3. b스레드가 ans ++를 하여 ans = 2가 되고 컨텍스트 스위칭을 한 후

4. 다시 a가 ans를 +1할때는 1에서 +1을 하는 것이기에 ans = 3이 되지 않고 ans = 2가 되는 경우가 생긴다.


따라서 이러한 Unsafe Region을 제거하기 위해 뮤텍스를 이용해본다.


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
#include <iostream>
#include <cstdio>
#include <queue>
#include <thread>
#include <mutex>
 
using namespace std;
 
int ans;
mutex myMutex;
void func() {
    myMutex.lock();
    for (int i = 1; i <= 1000000; i++)
        ans++;
    myMutex.unlock();
}
 
int main() {
    thread a(func);
    thread b(func);
    thread c(func);
 
    a.join();
    b.join();
    c.join();
    
    cout << "최종 ans :: " << ans << endl;
    return 0;
}
 
//                                                       This source code Copyright belongs to Crocus
//                                                        If you want to see more? click here >>
Crocus


이 과정에서 a가 myMutex.lock()를 하게 되면 b,c는 락이 걸려있기에 접근 할 수 없게되고, a가 끝나야 b 또는 c가 참여할 수 있게 된다.


따라서 mutex같은 것을 이용하여 비동기화를 동기화 시킬 수 있다. 


(이때 join을 하는 이유는 main 스레드가 끝나면 프로그램이 종료되는데 join을 이용하면 스레드가 종료 될 때 까지 메인 스레드가 기다려주는 역할을 하게 된다.)




반응형