JVM内部 - 为什么同一段代码在不同的时间运行？

Question

I am trying to understand the JVM options and came across the option CompileThreshold . 我试图了解JVM选项并遇到了选项CompileThreshold 。 I am running the below program: 我正在运行以下程序：

public class Main {    
    public static void main (String [] args) {    
        for (int repeat = 0; repeat < 200; ++ repeat) {    
            long start = System.nanoTime();    
            sum (100);    
            long end = System.nanoTime();    
            long diff = end - start;    
            System.out.println (repeat + ": " + diff);    
        }    
    }    `

    public static int sum (int n) {
        if (n <= 1)
            return 1;
        else
            return n + sum (n - 1);
    }
}

I get the below output: 我得到以下输出：

If you observe, there are spikes in the times at 32, 99, 149, 194. 如果你观察到，在32,99,149,194的时间里会出现尖峰。

After 194, the time taken to run is comparitively less. 在194之后，运行时间相对较少。 Please explain this behaviour. 请解释一下这种行为。 Thanks. 谢谢。

Answer 1

最初的尖峰是JIT决定编译代码，其余的是GC运行。

Answer 2

Most likly HotSpot in JVM re-JITs method after that many calls causing speedup after 194 run: http://java.sun.com/docs/books/performance/1st_edition/html/JPAppHotspot.fm.html 最常见的是HotVM在JVM重新JITs方法之后，在194次运行后导致加速的许多调用： http ： //java.sun.com/docs/books/performance/1st_edition/html/JPAppHotspot.fm.html

B.5.3 CompileThreshold B.5.3 CompileThreshold

Default Value: 1500 默认值：1500

Example Usage: java -XX:CompileThreshold=1000000 示例用法：java -XX：CompileThreshold = 1000000

The current implementation of HotSpot usually waits for a method to be executed a certain number of times before it is compiled. HotSpot的当前实现通常在编译之前等待一定数量的方法执行。 Not compiling every method helps startup time and reduces RAM footprint. 不编译每种方法都有助于启动时间并减少RAM占用空间。 This option allows you to control that threshold. 此选项允许您控制该阈值。 By increasing the number, you can trade slight reductions in RAM footprint in exchange for a longer period of time before your program reaches peak performance. 通过增加数量，您可以略微减少RAM占用空间，以换取更长的时间，然后程序达到最佳性能。

Answer 3

IMHO, this question is incomplete unless you mention the activity of the other processes running in the background/foreground ie without understanding the overall system load, it's kind of difficult to predict the reason for those timings. 恕我直言，这个问题是不完整的，除非你提到在后台/前台运行的其他进程的活动，即在不了解整体系统负载的情况下，很难预测这些时间的原因。 When running such benchmarks, make sure you capture/monitor the entire system load or at least make sure no background activities are running ie virus scans etc. Plus you are missing out other might-be-vital information like the JVM version, the platform and the number of cores. 运行此类基准测试时，请确保捕获/监视整个系统负载，或者至少确保没有正在运行的后台活动，即病毒扫描等。此外，您还缺少其他可能至关重要的信息，如JVM版本，平台和核心数量。

Would it be possible to confirm the same and re-post your results? 是否可以确认相同并重新发布结果？

JVM内部 - 为什么同一段代码在不同的时间运行？

问题描述

3 个解决方案

解决方案1
2 2012-06-20 05:11:41

解决方案2
1 2012-06-20 05:19:58

解决方案3
0 2012-06-20 05:26:35

JVM内部 -​​ 为什么同一段代码在不同的时间运行？

问题描述

3 个解决方案

解决方案1 2 2012-06-20 05:11:41

解决方案2 1 2012-06-20 05:19:58

解决方案3 0 2012-06-20 05:26:35

JVM内部 - 为什么同一段代码在不同的时间运行？

解决方案1
2 2012-06-20 05:11:41

解决方案2
1 2012-06-20 05:19:58

解决方案3
0 2012-06-20 05:26:35