在下面的示例代码中，为什么要使用join()方法？它的作用是什么？

在下面的示例代码中，为什么要使用join()方法？它的作用是什么？

package learning.multithreading;
import java.util.Random;import java.util.concurrent.ExecutionException;import java.util.concurrent.ForkJoinPool;import java.util.concurrent.RecursiveTask;
public class ParallelSumComputationUsingForkJoin {    private static final int[] LARGE_ARR = largeArr();
    private static final int LENGTH = LARGE_ARR.length;
    public static void main(String[] args) throws ExecutionException, InterruptedException {        RecursiveSumTask recursiveTask = new RecursiveSumTask(0, LENGTH, LARGE_ARR);        ForkJoinPool forkJoinPool = ForkJoinPool.commonPool();        long start = System.currentTimeMillis();        long sum = forkJoinPool.invoke(recursiveTask);        System.out.println("The sum is : "                + sum                + ", Time Taken by Parallel(Fork/Join) Execution: "                + (System.currentTimeMillis() - start) + " millis");    }
    private static int[] largeArr() {        return new Random().ints(500000000, 10, 1000).toArray();    }
    static class RecursiveSumTask extends RecursiveTask<Long> {
        private static final int SEQUENTIAL_COMPUTE_THRESHOLD = 4000;        private final int startIndex;        private final int endIndex;        private final int[] data;
        RecursiveSumTask(int startIndex, int endIndex, int[] data) {            this.startIndex = startIndex;            this.endIndex = endIndex;            this.data = data;        }
        @Override        protected Long compute() {            if (SEQUENTIAL_COMPUTE_THRESHOLD >= (endIndex - startIndex)) {                long sum = 0;                for (int i = startIndex; i < endIndex; i++) {                    sum += data[i];                }                return sum;            }            int mid = startIndex + (endIndex - startIndex) / 2;            RecursiveSumTask leftSumTask = new RecursiveSumTask(startIndex, mid, data);            RecursiveSumTask rightSumTask = new RecursiveSumTask(mid, endIndex, data);            leftSumTask.fork(); // Fork the Left Task in a Separate Execution            long rightSum = rightSumTask.compute(); // Compute the Right Part            long leftSum = leftSumTask.join(); // Wait for the results from the Left Part            return leftSum + rightSum; // Return Both        }    }}/** * Output: * The sum is : 252235235953, Time Taken by Parallel(Fork/Join) Execution: 139 millis *