Java ForkJoinPool 实现异步计算

前言

假设我们要计算一个未经优化的 Fibonacci 数
可以简单的写出以下代码:

    static long fib(long n) {
        if (n <= 2)
            return 1L;
        return fib(n - 1) + fib(n - 2);
    }

使用ForkJoinPool

class Fib extends RecursiveTask<Long> {
    private int n;

    /**
     * @param n
     */
    public Fib(int n) {
        this.n = n;
    }

    @Override
    protected Long compute() {
        if (n <= 2)
            return 1L;
        else {
            Fib f1 = new Fib(n - 1);
            Fib f2 = new Fib(n - 2);
            invokeAll(f1, f2);
            return f1.join() + f2.join();
        }
    }
}

二者比较

import java.util.concurrent.ForkJoinPool;
import java.util.concurrent.RecursiveTask;

public class ForkJoinFib {
    public static void main(String[] args) {
        final int N = 34;
        int times = 100;
        ForkJoinPool fjp = new ForkJoinPool();
        System.out.println("+----耗时测试----+");
        long startTime = System.currentTimeMillis();
        while (times-- != 0) {
            Fib f = new Fib(N);
            fjp.invoke(f);
        }
        System.out.println("用时: " + (System.currentTimeMillis() - startTime) + "毫秒");

        times = 100;
        System.out.println("+----耗时测试----+");
        startTime = System.currentTimeMillis();
        while (times-- != 0) {
            fib(N);
        }
        System.out.println("用时: " + (System.currentTimeMillis() - startTime) + "毫秒");
    }

    static long fib(long n) {
        if (n <= 2)
            return 1L;
        return fib(n - 1) + fib(n - 2);
    }

}

class Fib extends RecursiveTask<Long> {
    private int n;

    /**
     * @param n
     */
    public Fib(int n) {
        this.n = n;
    }

    @Override
    protected Long compute() {
        if (n <= 2)
            return 1L;
        else {
            Fib f1 = new Fib(n - 1);
            Fib f2 = new Fib(n - 2);
            invokeAll(f1, f2);
            return f1.join() + f2.join();
        }
    }

}

结果。。。
可能是创建对象那里时间用多了。。。
🤣