PooledHeapByteBuf,带有池的堆内buffer,顾名思义,肯定比一般通过new出来的buffer性能好。把对象放入对象池缓存起来,一般都是因为创建该对象开销比较大,常见的有线程池(ThreadPool)、连接池(ConnectionPool)等。
PooledHeapByteBuf继承关系如下:
PooledHeapByteBuf –》 PooledByteBuf –》 AbstractReferenceCountedByteBuf –》 AbstractByteBuf –》 ByteBuf。
继承关系比较简单清晰。
先介绍几个相关的类:
PooledByteBufAllocator:buffer分配器,用来分配buffer(包括堆内和堆外)。
PoolArena:一块逻辑上的内存池,用来管理和组织buffer的,内部数据结构较复杂。
FastThreadLocal:较快的ThreadLocal(相对于jdk自带的),实现:线程T扩展于FastThreadLocalAccess,InternalThreadLocalMap是它的成员变量,set()时放入InternalThreadLocalMap的成员变量数组,下标是index,get()时从InternalThreadLocalMap的成员变量数组中下标是index处取。
Recycler:回收器,基于FastThreadLocal,缓存的本地对象是个stack,stack里装的是Handler,handler里装的buffer。新建buffer时,从Recycler中的FastThreadLocal取,FastThreadLocal取出当前线程的本地变量stack,stack.pop()弹出handler,handler调用newObject()生成。
练习代码debug下:
<dependency> <groupId>io.netty</groupId> <artifactId>netty-all</artifactId> <version>5.0.0.Alpha2</version> </dependency>
package com.zoo; import io.netty.buffer.ByteBuf; import io.netty.buffer.PooledByteBufAllocator; import io.netty.util.concurrent.FastThreadLocalThread; /** * * @author yankai913@gmail.com * @date 2015-3-31 */ public class PooledHeapByteBufTest { static PooledByteBufAllocator alloc = new PooledByteBufAllocator(); static void singleThread() { new FastThreadLocalThread(new Runnable() { @Override public void run() { for (int i = 0; i < 3; i++) { String name = Thread.currentThread().getName(); long start = System.currentTimeMillis(); ByteBuf buf = alloc.buffer(102400); long end = System.currentTimeMillis(); System.out.println(name + "\t" + buf + ", time:" + (end - start)); } } }).start(); } public static void main(String[] args) throws Exception { singleThread(); } }
打印结果:
Thread-1 SimpleLeakAwareByteBuf(PooledHeapByteBuf(ridx: 0, widx: 0, cap: 102400)), time:62
Thread-1 PooledHeapByteBuf(ridx: 0, widx: 0, cap: 102400), time:0
Thread-1 PooledHeapByteBuf(ridx: 0, widx: 0, cap: 102400), time:0