https://blog.csdn.net/hitwengqi/article/details/8015646
先是c++11之前的
1.最基礎,進程同時創建5個線程,各自調用同一個函數
#include <iostream> #include <pthread.h> //多線程相關操作頭文件,可移植眾多平台 using namespace std; #define NUM_THREADS 5 //線程數 void* say_hello( void* args ) { cout << "hello..." << endl; } //函數返回的是函數指針,便於后面作為參數 int main() { pthread_t tids[NUM_THREADS]; //線程id for( int i = 0; i < NUM_THREADS; ++i ) { int ret = pthread_create( &tids[i], NULL, say_hello, NULL ); //參數:創建的線程id,線程參數,線程運行函數的起始地址,運行函數的參數 if( ret != 0 ) //創建線程成功返回0 { cout << "pthread_create error:error_code=" << ret << endl; } } pthread_exit( NULL ); //等待各個線程退出后,進程才結束,否則進程強制結束,線程處於未終止的狀態 }
輸入命令:g++ -o muti_thread_test_1 muti_thread_test_1.cpp -lpthread linux下編譯。
wq@wq-desktop:~/coding/muti_thread$ ./muti_thread_test_1
hello...hello...
hello...
hello...
hello...
運行結果運行順序是亂的。
2.線程調用到函數在一個類中,那必須將該函數聲明為靜態函數函數
因為靜態成員函數屬於靜態全局區,線程可以共享這個區域,故可以各自調用。
#include <iostream> #include <pthread.h> using namespace std; #define NUM_THREADS 5 class Hello { public: static void* say_hello( void* args ) { cout << "hello..." << endl; } }; int main() { pthread_t tids[NUM_THREADS]; for( int i = 0; i < NUM_THREADS; ++i ) { int ret = pthread_create( &tids[i], NULL, Hello::say_hello, NULL ); if( ret != 0 ) { cout << "pthread_create error:error_code" << ret << endl; } } pthread_exit( NULL ); }
wq@wq-desktop:~/coding/muti_thread$ ./muti_thread_test_2
hello...
hello...
hello...
hello...
hello...
順序也是亂的。
3.如何在線程調用函數時傳入參數呢? 先看下面修改的代碼,傳入線程編號作為參數:
#include <iostream> #include <pthread.h> //多線程相關操作頭文件,可移植眾多平台 using namespace std; #define NUM_THREADS 5 //線程數 void* say_hello( void* args ) { int i = *( (int*)args ); //對傳入的參數進行強制類型轉換,由無類型指針轉變為整形指針,再用*讀取其指向到內容 cout << "hello in " << i << endl; } //函數返回的是函數指針,便於后面作為參數 int main() { pthread_t tids[NUM_THREADS]; //線程id cout << "hello in main.." << endl; for( int i = 0; i < NUM_THREADS; ++i ) { int ret = pthread_create( &tids[i], NULL, say_hello, (void*)&i ); //傳入到參數必須強轉為void*類型,即無類型指針,&i表示取i的地址,即指向i的指針 cout << "Current pthread id = " << tids[i] << endl; //用tids數組打印創建的進程id信息 if( ret != 0 ) //創建線程成功返回0 { cout << "pthread_create error:error_code=" << ret << endl; } } pthread_exit( NULL ); //等待各個線程退出后,進程才結束,否則進程強制結束,線程處於未終止的狀態 }
wq@wq-desktop:~/coding/muti_thread$ ./muti_thread_test_3 hello in main.. Current pthread id = 3078458224 Current pthread id = 3070065520 hello in hello in 2 1 Current pthread id = hello in 2 3061672816 Current pthread id = 3053280112 hello in 4 Current pthread id = hello in 4 3044887408
顯然不是想要的結果,調用順序很亂,這是為什么呢?
這是因為多線程到緣故,主進程還沒開始對i賦值,線程已經開始跑了...?
#include <iostream> #include <pthread.h> //多線程相關操作頭文件,可移植眾多平台 using namespace std; #define NUM_THREADS 5 //線程數 void* say_hello( void* args ) { cout << "hello in thread " << *( (int *)args ) << endl; } //函數返回的是函數指針,便於后面作為參數 int main() { pthread_t tids[NUM_THREADS]; //線程id int indexes[NUM_THREADS]; //用來保存i的值避免被修改 for( int i = 0; i < NUM_THREADS; ++i ) { indexes[i] = i; int ret = pthread_create( &tids[i], NULL, say_hello, (void*)&(indexes[i]) ); if( ret != 0 ) //創建線程成功返回0 { cout << "pthread_create error:error_code=" << ret << endl; } } for( int i = 0; i < NUM_THREADS; ++i ) pthread_join( tids[i], NULL ); //pthread_join用來等待一個線程的結束,是一個線程阻塞的函數 }
wq@wq-desktop:~/coding/muti_thread$ ./muti_thread_test_3 hello in thread hello in thread hello in thread hello in thread hello in thread 30124
這是正常的嗎?感覺還是有問題...待續
代碼中如果沒有pthread_join主線程會很快結束從而使整個進程結束,從而使創建的線程沒有機會開始執行就結束了。加入pthread_join后,主線程會一直等待直到等待的線程結束自己才結束,使創建的線程有機會執行。
4.線程創建時屬性參數的設置pthread_attr_t及join功能的使用 線程的屬性由結構體pthread_attr_t進行管理。 typedef struct { int detachstate; 線程的分離狀態 int schedpolicy; 線程調度策略 struct sched_param schedparam; 線程的調度參數 int inheritsched; 線程的繼承性 int scope; 線程的作用域 size_t guardsize; 線程棧末尾的警戒緩沖區大小 int stackaddr_set; void * stackaddr; 線程棧的位置 size_t stacksize; 線程棧的大小 }pthread_attr_t;
#include <iostream> #include <pthread.h> using namespace std; #define NUM_THREADS 5 void* say_hello( void* args ) { cout << "hello in thread " << *(( int * )args) << endl; int status = 10 + *(( int * )args); //線程退出時添加退出的信息,status供主程序提取該線程的結束信息 pthread_exit( ( void* )status ); } int main() { pthread_t tids[NUM_THREADS]; int indexes[NUM_THREADS]; pthread_attr_t attr; //線程屬性結構體,創建線程時加入的參數 pthread_attr_init( &attr ); //初始化 pthread_attr_setdetachstate( &attr, PTHREAD_CREATE_JOINABLE ); //是設置你想要指定線程屬性參數,這個參數表明這個線程是可以join連接的,join功能表示主程序可以等線程結束后再去做某事,實現了主程序和線程同步功能 for( int i = 0; i < NUM_THREADS; ++i ) { indexes[i] = i; int ret = pthread_create( &tids[i], &attr, say_hello, ( void* )&( indexes[i] ) ); if( ret != 0 ) { cout << "pthread_create error:error_code=" << ret << endl; } } pthread_attr_destroy( &attr ); //釋放內存 void *status; for( int i = 0; i < NUM_THREADS; ++i ) { int ret = pthread_join( tids[i], &status ); //主程序join每個線程后取得每個線程的退出信息status if( ret != 0 ) { cout << "pthread_join error:error_code=" << ret << endl; } else { cout << "pthread_join get status:" << (long)status << endl; } } }
wq@wq-desktop:~/coding/muti_thread$ ./muti_thread_test_4 hello in thread hello in thread hello in thread hello in thread 0hello in thread 321 4 pthread_join get status:10 pthread_join get status:11 pthread_join get status:12 pthread_join get status:13 pthread_join get status:14
5.互斥鎖的實現
互斥鎖是實現線程同步的一種機制,只要在臨界區前后對資源加鎖就能阻塞其他進程的訪問。
#include <iostream> #include <pthread.h> using namespace std; #define NUM_THREADS 5 int sum = 0; //定義全局變量,讓所有線程同時寫,這樣就需要鎖機制 pthread_mutex_t sum_mutex; //互斥鎖 void* say_hello( void* args ) { cout << "hello in thread " << *(( int * )args) << endl; pthread_mutex_lock( &sum_mutex ); //先加鎖,再修改sum的值,鎖被占用就阻塞,直到拿到鎖再修改sum; cout << "before sum is " << sum << " in thread " << *( ( int* )args ) << endl; sum += *( ( int* )args ); cout << "after sum is " << sum << " in thread " << *( ( int* )args ) << endl; pthread_mutex_unlock( &sum_mutex ); //釋放鎖,供其他線程使用 pthread_exit( 0 ); } int main() { pthread_t tids[NUM_THREADS]; int indexes[NUM_THREADS]; pthread_attr_t attr; //線程屬性結構體,創建線程時加入的參數 pthread_attr_init( &attr ); //初始化 pthread_attr_setdetachstate( &attr, PTHREAD_CREATE_JOINABLE ); //是設置你想要指定線程屬性參數,這個參數表明這個線程是可以join連接的,join功能表示主程序可以等線程結束后再去做某事,實現了主程序和線程同步功能 pthread_mutex_init( &sum_mutex, NULL ); //對鎖進行初始化 for( int i = 0; i < NUM_THREADS; ++i ) { indexes[i] = i; int ret = pthread_create( &tids[i], &attr, say_hello, ( void* )&( indexes[i] ) ); //5個進程同時去修改sum if( ret != 0 ) { cout << "pthread_create error:error_code=" << ret << endl; } } pthread_attr_destroy( &attr ); //釋放內存 void *status; for( int i = 0; i < NUM_THREADS; ++i ) { int ret = pthread_join( tids[i], &status ); //主程序join每個線程后取得每個線程的退出信息status if( ret != 0 ) { cout << "pthread_join error:error_code=" << ret << endl; } } cout << "finally sum is " << sum << endl; pthread_mutex_destroy( &sum_mutex ); //注銷鎖 }
wq@wq-desktop:~/coding/muti_thread$ ./muti_thread_test_5 hello in thread hello in thread hello in thread 410 before sum is hello in thread 0 in thread 4 after sum is 4 in thread 4hello in thread 2 3 before sum is 4 in thread 1 after sum is 5 in thread 1 before sum is 5 in thread 0 after sum is 5 in thread 0 before sum is 5 in thread 2 after sum is 7 in thread 2 before sum is 7 in thread 3 after sum is 10 in thread 3 finally sum is 10
可知,sum的訪問和修改順序是正常的,這就達到了多線程的目的了,但是線程的運行順序是混亂的,混亂就是正常?
6.信號量的實現
信號量是線程同步的另一種實現機制,信號量的操作有signal和wait,本例子采用條件信號變量pthread_cond_t tasks_cond;
信號量的實現也要給予鎖機制。
#include <iostream> #include <pthread.h> #include <stdio.h> using namespace std; #define BOUNDARY 5 int tasks = 10; pthread_mutex_t tasks_mutex; //互斥鎖 pthread_cond_t tasks_cond; //條件信號變量,處理兩個線程間的條件關系,當task>5,hello2處理,反之hello1處理,直到task減為0 void* say_hello2( void* args ) { pthread_t pid = pthread_self(); //獲取當前線程id cout << "[" << pid << "] hello in thread " << *( ( int* )args ) << endl; bool is_signaled = false; //sign while(1) { pthread_mutex_lock( &tasks_mutex ); //加鎖 if( tasks > BOUNDARY ) { cout << "[" << pid << "] take task: " << tasks << " in thread " << *( (int*)args ) << endl; --tasks; //modify } else if( !is_signaled ) { cout << "[" << pid << "] pthread_cond_signal in thread " << *( ( int* )args ) << endl; pthread_cond_signal( &tasks_cond ); //signal:向hello1發送信號,表明已經>5 is_signaled = true; //表明信號已發送,退出此線程 } pthread_mutex_unlock( &tasks_mutex ); //解鎖 if( tasks == 0 ) break; } } void* say_hello1( void* args ) { pthread_t pid = pthread_self(); //獲取當前線程id cout << "[" << pid << "] hello in thread " << *( ( int* )args ) << endl; while(1) { pthread_mutex_lock( &tasks_mutex ); //加鎖 if( tasks > BOUNDARY ) { cout << "[" << pid << "] pthread_cond_signal in thread " << *( ( int* )args ) << endl; pthread_cond_wait( &tasks_cond, &tasks_mutex ); //wait:等待信號量生效,接收到信號,向hello2發出信號,跳出wait,執行后續 } else { cout << "[" << pid << "] take task: " << tasks << " in thread " << *( (int*)args ) << endl; --tasks; } pthread_mutex_unlock( &tasks_mutex ); //解鎖 if( tasks == 0 ) break; } } int main() { pthread_attr_t attr; //線程屬性結構體,創建線程時加入的參數 pthread_attr_init( &attr ); //初始化 pthread_attr_setdetachstate( &attr, PTHREAD_CREATE_JOINABLE ); //是設置你想要指定線程屬性參數,這個參數表明這個線程是可以join連接的,join功能表示主程序可以等線程結束后再去做某事,實現了主程序和線程同步功能 pthread_cond_init( &tasks_cond, NULL ); //初始化條件信號量 pthread_mutex_init( &tasks_mutex, NULL ); //初始化互斥量 pthread_t tid1, tid2; //保存兩個線程id int index1 = 1; int ret = pthread_create( &tid1, &attr, say_hello1, ( void* )&index1 ); if( ret != 0 ) { cout << "pthread_create error:error_code=" << ret << endl; } int index2 = 2; ret = pthread_create( &tid2, &attr, say_hello2, ( void* )&index2 ); if( ret != 0 ) { cout << "pthread_create error:error_code=" << ret << endl; } pthread_join( tid1, NULL ); //連接兩個線程 pthread_join( tid2, NULL ); pthread_attr_destroy( &attr ); //釋放內存 pthread_mutex_destroy( &tasks_mutex ); //注銷鎖 pthread_cond_destroy( &tasks_cond ); //正常退出 }
測試結果:
先在線程2中執行say_hello2,再跳轉到線程1中執行say_hello1,直到tasks減到0為止。
wq@wq-desktop:~/coding/muti_thread$ ./muti_thread_test_6 [3069823856] hello in thread 2 [3078216560] hello in thread 1[3069823856] take task: 10 in thread 2 [3069823856] take task: 9 in thread 2 [3069823856] take task: 8 in thread 2 [3069823856] take task: 7 in thread 2 [3069823856] take task: 6 in thread 2 [3069823856] pthread_cond_signal in thread 2 [3078216560] take task: 5 in thread 1 [3078216560] take task: 4 in thread 1 [3078216560] take task: 3 in thread 1 [3078216560] take task: 2 in thread 1 [3078216560] take task: 1 in thread 1
傳多個指針的
涉及多參數傳遞給線程的,都需要使用結構體將參數封裝后,將結構體指針傳給線程 定義一個結構體
struct mypara { var para1;//參數1 var para2;//參數2 }
將這個結構體指針,作為void *形參的實際參數傳遞 struct mypara pstru; pthread_create(&ntid, NULL, thr_fn,& (pstru));
函數中需要定義一個mypara類型的結構指針來引用這個參數 void *thr_fn(void *arg) { mypara *pstru; pstru = (struct mypara *) arg; pstru->para1;//參數1 pstru->para2;//參數2 }
pthread_create函數接受的參數只有一個void *型的指針,這就意味着你只能通過結構體封裝超過一個以上的參數作為一個整體傳遞。這是pthread_create函數的接口限定的,別人已經明確表明 我只接受一個參數,你硬要塞給他兩個肯定會出錯了。所以通過結構體這種組合結構變通一下,同樣實現了只通過一個參數傳遞,但通過結構指針對結構數據成員的 引用實現多參數的傳遞 這種用結構體封裝多參數的用法不僅僅用在pthread_create函數中,如果你自己設計的函數需要的參數很多〉=5個以上,都可以考慮使用結構體封 裝,這樣對外你的接口很簡潔清晰,你的函數的消費者使用起來也很方便,只需要對結構體各個成員賦值即可,避免了參數很多時漏傳、誤傳(參數串位)的問題
https://blog.csdn.net/ouyangfushu/article/details/80199140
c++11之后
C++多線程類Thread(C++11)
#include <thread>
#include <condition_variable>
#include <mutex>
C++11中std命名空間將Boost庫中的Thread加入,Boost的多線程從准標准變為標准,在深度學習以及應用程序開發中經常用到多線程,這里將其用法整理復習,以demo的形式復習,每次遇到問題或者忘記了總是百度,用完了就是忘記,好記性不如爛筆頭,Go ahead!
頭文件為#include<thread>,通過std::thread應用。就以Hello thread開始吧,需要注意的是join()函數和detach()函數的區別,數據同步操作mutex(需包含include<mutex>):互斥鎖
1、 普通函數多線程調用
thread.get_id()可以獲得線程的id.
#include <iostream> #include <thread> std::thread::id main_thread_id = std::this_thread::get_id(); void hello() { std::cout << "Hello Concurrent World\n"; if (main_thread_id == std::this_thread::get_id()) std::cout << "This is the main thread.\n"; else std::cout << "This is not the main thread.\n"; } void pause_thread(int n) { std::this_thread::sleep_for(std::chrono::seconds(n)); std::cout << "pause of " << n << " seconds ended\n"; } int main() { std::thread t(hello); std::cout << t.hardware_concurrency() << std::endl;//可以並發執行多少個(不准確) std::cout << "native_handle " << t.native_handle() << std::endl;//可以並發執行多少個(不准確) t.join(); std::thread a(hello); a.detach(); std::thread threads[5]; // 默認構造線程 std::cout << "Spawning 5 threads...\n"; for (int i = 0; i < 5; ++i) threads[i] = std::thread(pause_thread, i + 1); // move-assign threads std::cout << "Done spawning threads. Now waiting for them to join:\n"; for (auto &thread : threads) thread.join(); std::cout << "All threads joined!\n"; }
(1)無參數函數
(2)有參數函數
2、 在類內部創建線程
(1)類內部函數為靜態函數
在這里start()和hellothread()方法都必須是static方法。
(2)在Singleton模式內部創建線程:
3 、用類內部函數在類外部創建線程:
非常普通的類,只是用多線程調用其內部的函數
4、 join()和detach()的區別:
join()的作用前面已經提到,主線程等待子線程結束方可執行下一步(串行),detach()是的子線程放飛自我,獨立於主線程並發執行,主線程后續代碼段無需等待。看看效果:
(1)join()
(2)detach()
5、 數據同步(線程同時操作一個數據的安全性):