我正在试图弄清楚多线程是如何工作的,下面是我的代码:
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <fcntl.h>
#include <string.h>
#include <pthread.h>
static pthread_cond_t threadDied = PTHREAD_COND_INITIALIZER ; // cond var initialization
static pthread_mutex_t threadMutex = PTHREAD_MUTEX_INITIALIZER ; // mutex initialization
// this mutex will protect all of the below global vars
static int totThreads = 0 ; // total number of threads created
static int numLive = 0 ; // Total no. of threads still alive .. or terminated but not joined
static int numUnjoined = 0 ; // no. of threads that have not yet been joined
enum tstate { // enumeration of thread states
TS_ALIVE, // thread is alive
TS_TERMINATED, // thread terminated, not yet joined
TS_JOINED // thread terminated and joined
};
static struct { // info about each thread
pthread_t tid ; // thread ID
enum tstate state; // Thread state as per the above enum
int sleepTime ; // no. of seconds to live before terminating
} *thread ; // name of the struct .. well a pointer
static void *threadFunc (void *arg) { // default start function for each thread
int idx = *(int *)arg ; // since arg is of type void , we typecast it to * of type int and deref it
int s ; // for ret val
sleep(thread[idx].sleepTime) ; // pretending as though thread is doing some work :/
s = pthread_mutex_lock(&threadMutex);
if (s!=0) {
printf("whoops, couldn't acquire mutex\n") ;
fflush(stdout);
exit (-1) ;
}
numUnjoined ++ ;
thread[idx].state = TS_TERMINATED ;
s = pthread_mutex_unlock(&threadMutex) ;
if ( s!=0 ) {
printf("whoops, couldn't release mutex\n") ;
fflush(stdout);
exit (-2) ;
}
s = pthread_cond_signal(&threadDied) ; // signalling any listening thread to wake up !!
if (s != 0) {
printf("whoops, couldn't signal the main thread to reap\n");
fflush(stdout);
exit (-3) ;
}
printf("Thread %d has worked hard and is now terminating\n", idx);
fflush(stdout);
return NULL ;
}
int main(int argc, char *argv[]) {
int s, idx ;
if (argc < 2 || strcmp(argv[1], "--help") == 0) {
printf("Usage : %s nsecs...\n", argv[0]);
fflush(stdout);
exit(-4) ;
}
thread = calloc(argc -1, sizeof(*thread) );
if (thread == NULL) {
printf("whoops, couldn't allocate memory of size %lu\n", (argc -1) * sizeof(*thread) );
fflush(stdout);
exit(-5);
}
// Let's create all the threads now !!
for (idx =0 ; idx < argc -1 ; idx++ ) {
thread[idx].sleepTime = atoi(argv[idx + 1 ]) ; // thread sleeps for the duration entered in the cmd line
thread[idx].state = TS_ALIVE ;
s = pthread_create(&thread[idx].tid, NULL, threadFunc, &idx);
printf("Main created thread %d with tid : %lu \n", ( * (int *)&idx ), (unsigned long)thread[idx].tid);
fflush(stdout);
if (s != 0 ){
printf("whoops couldn't create thread %lu\n",(unsigned long) (&thread[idx].tid) );
fflush(stdout);
exit(-6) ;
}
//sleep(1); // << -- if I don't add this sleep, then it just deadlocks
}
totThreads = argc -1 ;
numLive = totThreads ;
// Join terminated threads
while (numLive > 0 ) {
s = pthread_mutex_lock(&threadMutex) ;
if (s!=0){
printf("whoops, couldn't lock mutex for joining\n") ;
fflush(stdout);
exit(-7) ;
}
while (numUnjoined == 0) {
s = pthread_cond_wait(&threadDied, &threadMutex) ;
if (s!=0) {
printf("whoops, couldn't wait for thread join\n") ;
fflush(stdout);
exit(-8) ;
}
}
for (idx = 0 ; idx < totThreads ; idx++ ) {
if (thread[idx].state == TS_TERMINATED) {
s = pthread_join(thread[idx].tid, NULL) ;
if (s!=0) {
printf("Failed thread join\n");
fflush(stdout);
exit(-9) ;
}
thread[idx].state = TS_JOINED ;
numLive-- ;
numUnjoined-- ;
printf("Reaped thread %d (numLive=%d)\n", idx, numLive);
fflush(stdout);
}
}
s = pthread_mutex_unlock(&threadMutex) ;
if (s!=0){
printf("whopps, couldn't unlock mutex after joining\n");
fflush(stdout);
exit(-10) ;
}
}
exit(EXIT_SUCCESS);
}
对于线程数为1的情况,此代码有时可以工作,有时则挂起:(
工作:
#./thread\u多连接1
主创建的线程0,tid为139835063281408
线程0工作正常,现在正在终止
收割的线程0(numLive=0)
挂起:
#./thread\u多连接1
主创建的线程0,tid为140301613573888
线程1工作正常,现在正在终止
^C类
注意这里Main说“线程0已创建”;而线程本身表示“线程1”。。。为什么会出现不匹配??
当我有多个线程时,它肯定会卡住:
#./thread\u multijoin 1 2 2 1
主创建的线程0,tid为140259455936256
主创建线程1,tid为:140259447543552
主创建线程2,tid为140259439150848
主创建线程3,tid:140259430758144
线程4工作正常,现在正在终止
线程0工作正常,现在正在终止
收割螺纹0(numLive=3)
收割螺纹3(numLive=2)
线程3工作正常,现在正在终止
收割螺纹2(numLive=1)
线程2工作很努力,现在正在终止
^C类
我从中了解到的唯一一件事是,main报告的线程ID与线程本身不同,因此我猜测,由于并行调度,线程计数器出现了一些问题。。。你们能帮我缩小范围吗?
提前谢谢。
========================================
感谢@mevets和@user3386109的回答:)
我试着按照@mevets的建议去做:I,e
pthread\u创建(&thread[idx])。tid,NULL,threadFunc,(void*)idx);
和
int idx=(int)arg;
但在编译时出错:
thread_multijoin.c: In function âthreadFuncâ:
thread_multijoin.c:32:15: error: cast from pointer to integer of different
size [-Werror=pointer-to-int-cast]
int idx = (int)arg ; // since arg is of type void , we typecast it to * of type int and deref it
thread_multijoin.c: In function âmainâ:
thread_multijoin.c:90:64: error: cast to pointer from integer of different
size [-Werror=int-to-pointer-cast]
s = pthread_create(&thread[idx].tid, NULL, threadFunc, (void *)idx );
经进一步研究,发现以下线索:
cast to pointer from integer of different size, pthread code
建议使用intptr\t:
s = pthread_create(&thread[idx].tid, NULL, threadFunc, (void *)(intptr_t)idx );
和
int idx = (intptr_t)arg
这样做非常好,没有错误。再次感谢您的时间,非常感谢:)
PS:要使用intptr\u t,您需要使用\u GNU\u源:
#define _GNU_SOURCE
