Improve calculation time in parallel mode

Question

I am new in the multithreading in Java, so I have a question about how to reduce calculation time in this example, without using Executors, Frameworks, etc, only plain threads?

public static void main(String[] args) throws TestException {

    Set<Double> res = new HashSet<>();

    for (int i = 0; i < TestConsts.N; i++) {
        res.addAll(TestCalc.calculate(i));
    }

    System.out.println(res);
}

And there is calculation method:

private static final Random rnd = new Random();

public static Set<Double> calculate(int num) throws TestException {
    // Emulates calculation delay time.
    try {
        Thread.sleep(rnd.nextInt(1000) + 1);
    }
    catch (InterruptedException e) {
        throw new TestException("Execution error.", e);
    }

    Set<Double> res = new HashSet<>();

    int n = rnd.nextInt(num + 1) + 1;

    for (int j = 0; j < n; j++) {
        res.add(rnd.nextDouble());
    }

    return res;
}

The reason for not using any frameworks is that I want to understand the original multithreading.

Thanks you in advance.

Answer 1

I will try to answer on a conceptual level:

• You could spawn a Thread for each task
• You could spawn n threads that use a (synchronized) Queue<Task> to obtain more work

You will have to synchronize whenever a thread finishes its part.