使用 pthreads 进行练习,但我的代码中出现了一些意外行为

Exercising with pthreads but I am having some unexpected behaviours in my code

我想用 c 编写一个程序,它将使用 pthreads 对数组的平方数求和以实现代码的并行执行,但是在 linux 环境中执行代码时我得到分段错误(核心被愚弄) 消息。

奇怪的是,当我通过 DevC++ 在 windows 环境中 运行 完全相同的代码时,代码执行没有问题。

除此之外,似乎只有 1 个线程正在完成导致串行执行的工作。

我不知道我是否错了,但 pthread_self() 返回的是同一个 ID。我还是pthreads和一般编程的新手,我找不到问题所在,所以我需要你的帮助。

#include <stdio.h>
#include <semaphore.h>
#include <pthread.h>
#include <stdlib.h>
#include <math.h>

void *calculation(int table[], int k);


pthread_mutex_t sum_mutex = PTHREAD_MUTEX_INITIALIZER;
int  n, p, sum = 0;

int main()
{
    int i, k=0, check, *table;

    printf("Give the number of threads:\n");
    scanf("%d",&p);
    pthread_t threads[p];

    printf("Give the number of elements of the table:");
    do{
    printf("The number of elements must be an integral multiple of the number of threads\n");
    scanf("%d",&n);
    check=n%p;
    if(check!=0){
    printf("Jesus how hard is it?\n");
    printf("Try one more time\n");
    printf("Give the number of elements of the table:\n");
    }
    }while(check!=0);

    table = (int*) malloc(n * sizeof(int));
    if(table == NULL){          
        printf("Error! Memory not allocated.\n");
        exit(0);
    }

    printf("Give the elements of the table:\n");
    for(i=0;i<n;i++){
        scanf("%d",&table[i]);
    }

    for(i=0;i<p;i++){                                                       //thread creation
        pthread_create(&threads[i], NULL, calculation(table, k), NULL);
        k++;                                                                //k is a variable used to seperate table, explained later
    }
    for(i=0;i<p;i++){
        pthread_join(threads[i],NULL);
    }

    printf("Sum of vector= %d\n",sum);

    free(table);
    exit(0);
    return 0;
}

void *calculation(int table[], int k){
    int i;
    int local_sum=0;
    for(i=(n/p)*k;i<(n/p)*(k+1);i++)                                        //this algorithm seperates the table into equivelant pieces and
    {                                                                       //every pthread is calculating its given piece then stores that value in its local variable sum
        if((n/p)>n){                                                        //then it is updating the global variable
            pthread_exit(NULL);
        }
        local_sum+=pow(table[i], 2);
        printf("Thread's %lu calculation is %d\n", pthread_self(), local_sum);
    }

    pthread_mutex_lock(&sum_mutex);                                         //mutex used here to protect the critical code
    sum += local_sum;
    pthread_mutex_unlock(&sum_mutex);
}

如果我没记错的话一个线程是运行一个不同的"copy"代码,因此每个线程的局部变量是不同的。在计算 local_sum 之后,它会更新出于明显原因而受到保护的全局总和。

正如我所说,在 windows 环境中,这段代码 运行 很流畅,但似乎是同一个 pthread 完成了所有工作。相反,应该将工作分成与存在的线程一样多的部分。

不要忘记在 linux 环境中核心根本没有运行导致错误的事实:分段错误(核心变笨)。

来自手册页,pthread_create

的声明
int pthread_create(pthread_t *thread, const pthread_attr_t *attr,
                     void *(*start_routine) (void *), void *arg);

pthread_create的第三个参数是一个函数指针,它以void*为参数。 在你的情况下你有两个参数 void *calculation(int table[], int k);

您向 pthread_create 传递参数的方式也是错误的,它应该只是函数名而不是其参数。(阅读更多关于回调函数的信息)。最后的NULL应该是参数。在你的情况下你想传递 tablek 但它应该是一个指针。

pthread_create(&threads[i], NULL, calculation(table, k), NULL);

下面,我修改了代码,引入了一个新的struct thread_arg,其中包含tablek,这个struct会传给你的calculation 然后解包并正常使用。

#include <stdio.h>
#include <semaphore.h>
#include <pthread.h>
#include <stdlib.h>
#include <math.h>

struct thread_arg {
        int *table;
        int k;
};

void *calculation(void *);

pthread_mutex_t sum_mutex = PTHREAD_MUTEX_INITIALIZER;
int  n, p, sum = 0;

int main()
{
        int i, k=0, check, *table;

        printf("Give the number of threads:\n");
        scanf("%d",&p);
        pthread_t threads[p];

        printf("Give the number of elements of the table:");
        do{
                printf("The number of elements must be an integral multiple of the number of threads\n");
                scanf("%d",&n);
                check=n%p;
                if(check!=0){
                        printf("Jesus how hard is it?\n");
                        printf("Try one more time\n");
                        printf("Give the number of elements of the table:\n");
                }
        }while(check!=0);

        table = (int*) malloc(n * sizeof(int));
        if(table == NULL){
                printf("Error! Memory not allocated.\n");
                exit(0);
        }

        printf("Give the elements of the table:\n");
        for(i=0;i<n;i++){
                scanf("%d",&table[i]);
        }

        // edit: since I know the size allocated memory for th_args;
        struct thread_arg th_args[p];

        for(i=0;i<p;i++){                                                       //thread creation
                th_args[i].table = table;
                th_args[i].k = k;
                pthread_create(&threads[i], NULL, calculation, &th_args[i]);
                k++;                                                                //k is a variable used to seperate table, explained later
        }
        for(i=0;i<p;i++){
                pthread_join(threads[i],NULL);
        }

        printf("Sum of vector= %d\n",sum);

        free(table);
        exit(0);
        return 0;
}

void *calculation(void *data){
        int i;
        int local_sum=0;
        int *table;
        int k;

        struct thread_arg *th_args = data;
        k = th_args->k;
        table = th_args->table;

        printf("shubham: thread_spawned\n");
        for(i=(n/p)*k;i<(n/p)*(k+1);i++)                                        //this algorithm seperates the table into equivelant pieces and
        {                                                                       //every pthread is calculating its given piece then stores that value in its local variable sum
                if((n/p)>n){                                                        //then it is updating the global variable
                        pthread_exit(NULL);
                }
                local_sum+=pow(table[i], 2);
                printf("Thread's %lu calculation is %d\n", pthread_self(), local_sum);
        }

        pthread_mutex_lock(&sum_mutex);                                         //mutex used here to protect the critical code
        sum += local_sum;
        pthread_mutex_unlock(&sum_mutex);

        // edit: function returs NULL;
        return NULL;
}

Edit1:现在 calculation returns NULL,通过创建 p th_args

处理了 k++ 案例