如何使用P处理器计算整数数组的元素之和

Question

I have to add the elements of an array of integers using P processors. Below it's the code i've written so far. I have defined an array and a number of 4 processors to can make some tests.

#include <stdio.h>
#include <unistd.h>
#include <signal.h>
#include <sys/wait.h>
#define P 17
#define SIZE 153

static pid_t id[P];
static int elementsList[] = {1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,
            1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,
            1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,
            1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,
            1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,
            1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,
            1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,
            1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,
            1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17
            };

void getSubvectorSum(int elemList[], int start, int step, int size,int wPipe){
int sum = 0;
int i;
for(i = start; i < size; i += step){
    sum += elemList[i];
}
write(wPipe,&sum,sizeof sum);
}

int main(){
int fd[2];
int total = 0;
int result;
int nbytes;
int i;
pipe(fd);
for(i = 0; i < P; i++){
    id[i] = fork();
    if(id[i] == 0){
                    close(fd[0]);
        getSubvectorSum(elementsList,i,P,SIZE,fd[1]);
        exit(0);
    }
}
for(i = 0; i < P; i++){
    wait(NULL);
}
close(fd[1];
for(i = 0; i < P; i++){
    nbytes = read(fd[0],&result,sizeof result);
    if(nbytes > 0){
        printf("Something on the pipe.\n");
        total += result;

    } else {
        printf("Nothing on the pipe.\n");
    }
}   
for(i = 0; i < P; i++){
    kill(id[i],SIGKILL);
}
printf("total result: %d\n",total);
return 0;

}

Am I doing things right?

Answer 1

You program might deadlock with large number of processes: once the pipe buffer is full, children will block on write() while the parent will wait() . Drain the pipe in the parent without waiting for the child processes to complete.

Your program might read less bytes than written by children.

Check error conditions of system calls. It might make it easier to uncover bugs.

I'm not sure but you might need to retry wait() call on EINTR error; otherwise you'll create zombies.

You shouldn't need to kill .

Look at parallel-sum-fork.c and parallel-sum-openmp.c for comparison.