Thread.sleep()和TimeUnit.SECONDS.sleep()的区别与联系

刚看到TimeUnit.SECONDS.sleep()方法时觉得挺奇怪的，这里怎么也提供sleep方法？
public void sleep(long timeout) throws InterruptedException {
    if (timeout > 0) {
        long ms = toMillis(timeout);
        int ns = excessNanos(timeout, ms);
        Thread.sleep(ms, ns);
    }
}
结果一看源码，原来是对Thread.sleep方法的包装，实现是一样的，只是多了时间单位转换和验证，然而TimeUnit枚举成员的方法却提供更好的可读性，这可能就是当初创建TimeUnit时提供sleep方法的原因吧，大家都知道sleep方法很常用，但经常要使用一个常量保存sleep的时间，比如3秒钟，我们代码通常会这样写:

private final int SLEEP_TIME = 3 * 1000; //3 seconds
因为Thread.sleep方法参数接受的毫秒单位的数值，比较下面代码就知道TimeUnit枚举成员的sleep方法更优雅：
TimeUnit.MILLISECONDS.sleep(10);
TimeUnit.SECONDS.sleep(10);
TimeUnit.MINUTES.sleep(10);
Thread.sleep(10);
Thread.sleep(10*1000);
Thread.sleep(10*60*1000);
但使用TimeUnit枚举成员的sleep方法会不会带来性能损失了，毕竟增加了函数调用开销？
测试测试吧：
import java.util.concurrent.TimeUnit;
public class TestSleep {
    public static void main(String[] args) throws InterruptedException {     
        sleepByTimeunit(10000);
                                                                                                                  
        sleepByThread(10000);    
    }
    private static void sleepByTimeunit(int sleepTimes) throws InterruptedException {
        long start = System.currentTimeMillis();
                                                                                                                  
        for(int i=0; i<sleepTimes; i++){
            TimeUnit.MILLISECONDS.sleep(10);
        }
                                                                                                                  
        long end = System.currentTimeMillis();
                                                                                                                  
        System.out.println("Total time consumed by TimeUnit.MILLISECONDS.sleep : " + (end - start));
    }
                                                                                                              
    private static void sleepByThread(int sleepTimes) throws InterruptedException {
        long start = System.currentTimeMillis();
                                                                                                                  
        for(int i=0; i<sleepTimes; i++){
            Thread.sleep(10);
        }
                                                                                                                  
        long end = System.currentTimeMillis();
                                                                                                                  
        System.out.println("Total time consumed by Thread.sleep : " + (end - start));
    }
}
两次测试结果（Win7+4G+JDK7 测试期间计算资源充足）：
Total time consumed by TimeUnit.MILLISECONDS.sleep : 100068
Total time consumed by Thread.sleep : 100134
Difference : -- -66
Total time consumed by TimeUnit.MILLISECONDS.sleep : 100222
Total time consumed by Thread.sleep : 100077
Difference : -- +145
从结果可以看出10000次调用差异很小，甚至一次更快，不排除JVM进行了优化，如果忽略性能方面考虑，从可读性方面建议使用TimeUnit枚举成员的sleep方法。
另外TimeUnit是枚举实现一个很好的实例，Doug Lea太神了，佩服得五体投地！
出处:http://stevex.blog.51cto.com/4300375/1285767

参考：https://www.cnblogs.com/sanhuan/p/4826401.html