#include <stdio.h>
#include <omp.h>

static double num = 1;

double func1(){
    for (int i=0; i<100; i  ){
        num  = 1;
        for (int j=0; j<100; j  ){
            num -= 1;
            for (int k=0; k<100; k  ){
                num  = 1;
                if (num == 1000){
                    return num;
                }
            }
        }
    }
}

int main(){
    #pragma omp parallel num_threads(10)
    {
        double num2 = func1();
    }
    printf("%lf", num2);
    return 0;
}

/*
Compiles with: gcc fileName.c -o fileName -openmp
Executes with: ./fileName
// Note: in the terminal
*/

The above code when compiled gets stuck in an endless loop or something it does not exit. I tried "#pragma omp parallel for" but it throws out an error. I wonder if it works if I collapse the loop into one. Would really appreciate help. Thanks.`

CodePudding user response：

This is a no-show: all your threads try updating simultaneously the num global variable. This is a so-called "race condition". Not only the final result is unpredictable, but this slows down everything.

As a rule of thumb, in multithreaded code do not use global and/or static variables, unless you really know what you are doing.

CodePudding user response：

Your code has as race condition. Interesting, adding reduction( :num) to the parallel clause doesn't solve much. You break at 1000 but I get results (if I make num2 global to the main) that are approximately 1000 nthreads. Big question is: what is this supposed to do?

For the rest I have some comments: 1. It can not compile because num2 is local. 2. Your function generates a warning that there are code paths that return void. 3. Why do you return num when it's global anyway? 4. Do not use global variables. 5. Your printf is missing a newline.