JDK 8 lambda和并行流性能比现有的JDK 6/7差

Question

我比较了一个1000万整数数组中一个简单的“find max”逻辑的性能。简单的for循环比lambda和并行流版本执行得多（至少好10倍）。

有人可以帮我理解这种反直觉行为吗？我有一台安装了Windows 7专业版的四核英特尔i5处理器戴尔E5530笔记本电脑， 1.8.0_60 64位JVM 。

import java.util.ArrayList;
import java.util.List;
import java.util.Optional;
import java.util.Random;

public class MaxFind
{
  private static final int CAPACITY = 10_000_000;

  public static void main(String[] args) throws Exception
  {
    List<Integer> numbers = new ArrayList<Integer>(CAPACITY);
    Random rand = new Random();

    long start = System.currentTimeMillis();
    for(int i=0; i<CAPACITY; i++)
        numbers.add(rand.nextInt());
    long end = System.currentTimeMillis();
    System.out.println("size of numbers: " + numbers.size() + ", time taken to populate: " + (end-start) + " milliseconds");

    start = System.currentTimeMillis();
    int maxNum = Integer.MIN_VALUE;

    //imperative approach
    for(int i=0; i<numbers.size(); i++)
        maxNum = Integer.max(numbers.get(i), maxNum);
    end = System.currentTimeMillis();

    System.out.println("Time taken to find the max value " + maxNum + " using normal for loop: " + (end-start) + " milliseconds.");

    start = System.currentTimeMillis();
    //lambda, parallel stream
    Optional<Integer> max = numbers.parallelStream().reduce(Integer::max);
    end = System.currentTimeMillis();
    System.out.println("Time taken to find the max value " + max.get() + " using lambda with parallel stream 1: " + (end-start) + " milliseconds.");

    start = System.currentTimeMillis();
    maxNum = numbers.parallelStream().reduce(Integer.MIN_VALUE, (a, b) -> Integer.max(a, b));
    end = System.currentTimeMillis();
    System.out.println("Time taken to find the max value " + max.get() + " using lambda with parallel stream 2: " + (end-start) + " milliseconds.");        
   }
}