【发布时间】:2017-11-29 22:41:08
【问题描述】:
在此post 中使用挂钟时间为任务计时不正确。
“挂钟时间”是用户等待任务完成所经历的真实时间。在 Java 1.5 之前,测量挂钟时间是对 Java 任务进行基准测试的常规(也是唯一)方法。
要测量挂钟时间,请在任务前后致电
java.lang.System . currentTimeMillis()并获取差额。 该方法以毫秒(千分之一秒)为单位返回时间。挂钟时间受到系统上其他活动的强烈影响,例如后台进程、其他应用程序、磁盘或网络活动以及显示器更新。在某些系统上,例如 Windows,如果应用程序不在顶部,它将以较低的优先级运行并且需要更长的时间。除非您非常小心地使用未加载的系统并在大量测试中取平均值,否则所有这些都会使您的基准测试结果出现偏差。
如果您对用户的体验感兴趣,使用挂钟时间不一定是坏事。但是,很难获得一致的基准数字来揭示您自己的应用程序的问题。
如果我们使用getProcessCpuTime(),似乎我们测量了JVM的所有时间。
那么如何区分JVM固有线程(below)和Java程序线程(单线程或多线程),以便测量真正的Java程序线程CPU时间?
enum ThreadType { // **intrinsic threads**
vm_thread,
cgc_thread, // Concurrent GC thread
pgc_thread, // Parallel GC thread
java_thread, // Java, CodeCacheSweeper, JVMTIAgent and Service threads.
compiler_thread,
watcher_thread,
os_thread
};
在这个post,main_thread 似乎是一个单线程。 mutator 线程是由 JVM 创建的吗?
(src/share/vm/runtime/thread.cpp)
3191 // Attach the main thread to this os thread
3192 JavaThread* main_thread = new JavaThread();
3193 main_thread->set_thread_state(_thread_in_vm);
3194 // must do this before set_active_handles and initialize_thread_local_storage
3195 // Note: on solaris initialize_thread_local_storage() will (indirectly)
3196 // change the stack size recorded here to one based on the java thread
3197 // stacksize. This adjusted size is what is used to figure the placement
3198 // of the guard pages.
3199 main_thread->record_stack_base_and_size();
3200 main_thread->initialize_thread_local_storage();
如果 Java 程序是多线程的, 像下面的程序会怎样?
// Java code for thread creation by implementing
// the Runnable Interface
class MultithreadingDemo implements Runnable
{
public void run()
{
try
{
// Displaying the thread that is running
System.out.println ("Thread " +
Thread.currentThread().getId() +
" is running");
}
catch (Exception e)
{
// Throwing an exception
System.out.println ("Exception is caught");
}
}
}
int n = 8; // Number of threads
for (int i=0; i<8; i++)
{
Thread object = new Thread(new MultithreadingDemo());
object.start();
}
为什么 MXBeans 只返回 1 个主线程?
import java.lang.management.ManagementFactory;
import java.lang.management.ThreadInfo;
import java.lang.management.ThreadMXBean;
import java.util.ArrayDeque;
import java.util.LinkedList;
public class MultiThread {
/** Get CPU time in nanoseconds. */
public long getCpuTime( ) {
ThreadMXBean bean = ManagementFactory.getThreadMXBean( );
return bean.isCurrentThreadCpuTimeSupported( ) ?
bean.getCurrentThreadCpuTime( ) : 0L;
}
/** Get user time in nanoseconds. */
public long getUserTime( ) {
ThreadMXBean bean = ManagementFactory.getThreadMXBean( );
return bean.isCurrentThreadCpuTimeSupported( ) ?
bean.getCurrentThreadUserTime( ) : 0L;
}
public static void main(String[] args) {
// deque operation
//ArrayDeque<LinkedList<Integer>> deque = new ArrayDeque<LinkedList<Integer>>();
ArrayDeque<Byte[]> deque = new ArrayDeque<Byte[]>();
int objectSize = 1024;
int xxx = 0;
// expand old and young heap size
for(long i = 0; i < 1500; i++){
for(int j = 0; j < 300; j++){
deque.addLast(new Byte[objectSize]);
}
for(int m = 0; m < 65535; m++){
xxx += 1;
xxx -= 1;
}
for(int j = 0; j < 20; j++){
deque.removeLast();
}
}
for(long i = 0; i < 1500; i++){
for(int j = 0; j < 300; j++){
deque.addLast(new Byte[objectSize]);
}
for(int m = 0; m < 65535; m++){
xxx += 1;
xxx -= 1;
}
for(int j = 0; j < 300; j++){
deque.removeFirst();
}
}
int n = 8; // Number of threads
for (int i=0; i<8; i++)
{
Thread object = new Thread(new MultithreadingDemo());
object.start();
}
ThreadMXBean threadMXBean = ManagementFactory.getThreadMXBean();
//don't need monitor and synchronizer info,only get thread and stack info。
ThreadInfo[] threadInfos = threadMXBean.dumpAllThreads(false, false);
for (ThreadInfo threadInfo : threadInfos) {
String tName = threadInfo.getThreadName();
long tId = threadInfo.getThreadId();
System.out.println(tName + "," + tId);
}
}
}
【问题讨论】: