fork and join

Fork & Join

Multithread programming with fork & join is ‘very’ fun. From JDK 1.7 onward, a framework called ‘fork & join’ is added. This framework makes it possible to split a task into many units of tasks, each of which is processed concurrently by different threads.

RecursiveAction - task without any return value RecursiveTask - task with return value

What you need to do is to implement compute() method by inheriting certain classes as

public class SumJob extends RecursiveTask<Long> { 
    long start, end;
    SumJob(long start, long end){
        this.start = start;
        this.end = end;
    }
    @Override
    protected Long compute() {
        //Task is divided 
        long size = end - start + 1;
        if(size <= 10)
            return sum();

        long half = (start + end)/2;

        SumJob leftJob = new SumJob(start, half);
        SumJob rightJob = new SumJob(half+1, end);

        leftJob.fork(); //fork is asynchronous
        return rightJob.compute() + leftJob.join(); //join is synchronous

    }
    public long sum(){
        long ret = 0l;
        for(long i = start; i<=end; i++){
            ret += i;
        }
        return ret;
    }
}

public class ForkMain {
    static final ForkJoinPool forkJoinPool = new ForkJoinPool();
    public static void main(String[] args) {
        long startTime;
        long start= 1;
        long end = 1000000;
        long result = 0;

        SumJob sumJob = new SumJob(start, end);

        startTime = System.currentTimeMillis();
        result = forkJoinPool.invoke(sumJob);
        System.out.println(result + ", Time taken is " + (System.currentTimeMillis()-startTime));

        result = 0;
        startTime = System.currentTimeMillis();
        for(long i = start; i<=end; i++){
            result += i;
        }
        System.out.println(result + ", Time taken is " + (System.currentTimeMillis()-startTime));
    }
}

Console Output

500000500000, Time taken is 44
500000500000, Time taken is 4

Time taken for multithread operation is much longer than that of simple single thread. Hence you should always investigate if using multi thread programming will actually be beneficial for certain operations.