顾名思义,Netty中的FastThreadLocal要比JDK中的ThreadLocal更快,那FastThreadLocal是怎么做到的呢?
FastThreadLocal的使用
FastThreadLocal的使用方法和ThreadLocal基本一致,但有一点必须注意的是,FastThreadLocal必须在FastThreadLocalThread中才能体现出“Fast”的优势,在普通Thread中会退化回ThreadLocal。
import io.netty.util.concurrent.FastThreadLocal;
import io.netty.util.concurrent.FastThreadLocalThread;
public class FastThreadLocalTest {
private static FastThreadLocal<String> fastThreadLocal = new FastThreadLocal<>();
public static void main(String[] args) {
fastThreadLocal.set("Main Thread");
new FastThreadLocalThread(() -> {
fastThreadLocal.set("Thread 1");
System.out.println("Thread 1 => " + fastThreadLocal.get());
}).start();
new FastThreadLocalThread(() -> {
fastThreadLocal.set("Thread 2");
System.out.println("Thread 2 => " + fastThreadLocal.get());
}).start();
System.out.println("Main Thread => " + fastThreadLocal.get());
}
}
程序输出:
Thread 1 => Thread 1
Main Thread => Main Thread
Thread 2 => Thread 2
Process finished with exit code 0
FastThreadLocal的原理
ThreadLocal中通过一个Map来维护每一个线程中ThreadLocal到Object的映射,这个Map使用了线性探测法来解决哈希冲突,在最坏情况下会有O(n)的复杂度,而且计算哈希码的过程本身也需要耗时。
FastThreadLocal通过空间换时间的方式解决了上述问题。每一个FastThreadLocal都有一个全局唯一的index,该index唯一确定了FastThreadLocal在每一个线程中的Object数组中的下标,因此,可以在O(1)时间内进行get和set操作。但其缺点是浪费空间,例如,某个FastThreadLocal的index很大,而某个线程只使用了此FastThreadLocal,那该线程也必须开辟至少index的数组大小才行。
下面我们来看一下源码。
首先,我们看一下FastThreadLocalThread中相关的字段。
/**
* A special {@link Thread} that provides fast access to {@link FastThreadLocal} variables.
*/
public class FastThreadLocalThread extends Thread {
...
private InternalThreadLocalMap threadLocalMap;
...
}
由注释也可知,FastThreadLocalThread提供了一个对FastThreadLocal变量的快速访问,其核心字段是InternalThreadLocalMap类型的变量threadLocalMap。下面我们看一下InternalThreadLocalMap的数据结构。
InternalThreadLocalMap继承自UnpaddedInternalThreadLocalMap:
class UnpaddedInternalThreadLocalMap {
static final ThreadLocal<InternalThreadLocalMap> slowThreadLocalMap = new ThreadLocal<InternalThreadLocalMap>();
static final AtomicInteger nextIndex = new AtomicInteger();
/** Used by {@link FastThreadLocal} */
Object[] indexedVariables;
...
}
可见其也是用数组存放了相关变量,继续看InternalThreadLocalMap。先看get方法:
public static InternalThreadLocalMap get() {
Thread thread = Thread.currentThread();
if (thread instanceof FastThreadLocalThread) {
return fastGet((FastThreadLocalThread) thread);
} else {
return slowGet();
}
}
可见,如果线程类型是FastThreadLocalThread,则使用fastGet方法,否则则用slowGet方法。需要注意的是,方法的返回值是InternalThreadLocalMap,而非具体存放的元素值。下面是fastGet和slowGet的源码:
private static InternalThreadLocalMap fastGet(FastThreadLocalThread thread) {
InternalThreadLocalMap threadLocalMap = thread.threadLocalMap();
if (threadLocalMap == null) {
thread.setThreadLocalMap(threadLocalMap = new InternalThreadLocalMap());
}
return threadLocalMap;
}
private static InternalThreadLocalMap slowGet() {
ThreadLocal<InternalThreadLocalMap> slowThreadLocalMap = UnpaddedInternalThreadLocalMap.slowThreadLocalMap;
InternalThreadLocalMap ret = slowThreadLocalMap.get();
if (ret == null) {
ret = new InternalThreadLocalMap();
slowThreadLocalMap.set(ret);
}
return ret;
}
下面我们看一下FastThreadLocal的get方法:
/**
* Returns the current value for the current thread
*/
@SuppressWarnings("unchecked")
public final V get() {
InternalThreadLocalMap threadLocalMap = InternalThreadLocalMap.get();
Object v = threadLocalMap.indexedVariable(index);
if (v != InternalThreadLocalMap.UNSET) {
return (V) v;
}
return initialize(threadLocalMap);
}
不难看出,FastThreadLocal直接使用index从当前线程的threadLocalMap中取值。
关于set相关的内容和其它一些琐碎细节,读者若有兴趣可自行探索,只要理解了前文中所说的FastThreadLocal的设计理念,读懂源码将没什么困难。记住,源码之前,了无秘密。
每日学习笔记,写于2020-05-04 星期一