Multithreading in a for loop misses the first iteration

October 22, 2022

I am writing an application requiring a variable number of threads, which I would like to create within a vector over which I could iterate. The application consistently skips the first iteration of the for-loop but instead performs an unintended operation at the end of the loop. I would like not only to solve the issue but also to understand its origin.

Here is a toy example to recreate the behavior:

#include <iostream>
#include <mutex>
#include <thread>
#include <vector>

std::mutex my_mutex;

void print_func(int idx){
    std::lock_guard<std::mutex> f(my_mutex);
    std::cout << "Thread id: " << idx << std::endl;
}

int main() {

    const auto processor_count = 4;
    std::vector<std::thread> ThreadVector;

    for(int i = 0; i < processor_count; i++){
        ThreadVector.emplace_back([&](){print_func(i);});
    }

    for(auto& t: ThreadVector){
        t.join();
    }
    return 0;
}

Here’s the output:

Thread id: 1
Thread id: 2
Thread id: 3
Thread id: -214744752

Here’s the intended behavior:

Thread id: 0
Thread id: 1
Thread id: 2
Thread id: 3

>Solution :

Your lambda:

[&](){print_func(i);}

captures the int i by reference, so when the thread started, i is no longer the same value, and for the last iteration, i goes out of scope, therefore it has garbage value from the stack.

Capturing by value instead of reference will solve your problem: