关于System.arraycopy效率的质疑

[Benjamin_whx]

package Collections;

import java.util.Arrays;

/**
 * Created by piqiu on 1/16/16.
 */
public class CollectionsTest {

    public static void main(String[] args) {
        int source[] = new int[100000000];
        Arrays.fill(source, 250);
        int destinate[] = new int[100000000];

//        useForEach(source, 0, destinate, 0, 100000000);  // useForEach spend: 63 millSeconds
        useMemocpy(source, 0, destinate, 0, 100000000);    // useMemocpy spend: 76 millSeconds
    }

    private static void useForEach(int[] source, int startSourceIndex, int[] destinate, int startDesIndex, int length) {
        long startMillSeconds = System.currentTimeMillis();

        int desIndex = startDesIndex;
        for (int i = startSourceIndex; i < length; i++) {
            destinate[desIndex] = source[startSourceIndex];
            desIndex++;
        }

        long endMillSeconds = System.currentTimeMillis();
        System.out.println("useForEach spend: " + (endMillSeconds - startMillSeconds) + " millSeconds");
    }

    private static void useMemocpy(int[] source, int startSourceIndex, int[] destinate, int startDesIndex, int length) {
        long startMillSeconds = System.currentTimeMillis();
        System.arraycopy(source, startSourceIndex, destinate, startDesIndex, length);
        long endMillSeconds = System.currentTimeMillis();
        System.out.println("useMemocpy spend: " + (endMillSeconds - startMillSeconds) + " millSeconds");
    }
}

上面的代码我是注释掉一个个运行的，发现数组长度10000000的时候，使用自己遍历来赋值花费5毫秒，使用System.arraycopy花费2毫秒，提升不大。然后我就把数组长度加了一个0，变成1000000000，突然System.arraycopy花费的时间比自己遍历还慢了？这是为什么呢？？？

[Benjamin_whx]

不是System.arraycopy效率很高，使用的是内存直接copy的吗？还有是最近java优化了吗，之前for循环没这么效率的啊。

[tanguojingzh]

你把内容换成中文，你就知道了。英文数字是特殊的