{"type":"doc","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"大多数程序员在日常开发中常常会碰到GC的问题：OOM异常、GC停顿等，这些异常直接导致糟糕的用户体验，如果不能得到及时处理，还会严重影响应用程序的性能。本系列从GC的基础入手，逐步帮助读者熟悉GC各种技术及问题根源。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"GC的由来"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"想当初，盘古开天辟地......"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"好吧，扯远了，这也不是仙侠小说..."}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"GC到底是怎么来的呢？这个问题要从C语言聊起， 大家都知道， C\/C++语言在编写程序的时候， 需要码神们自己管理内存， 直观的说就是使用内存的时候要malloc，之后这段内存会一直保留给程序进程，直到程序显式的调用free才会得以释放。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"一个例子引发的问题"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"\/\/第 0 步: char* aMem;\n\/\/第 1 步:\naMem = (char*) malloc(sizeof(char) * 1024);\n\/\/第 2 步:\nstrcpy(aMem, \"I am a bunch of memory\");\n\/\/第 3 步:\nfree(aMem);"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"看到没有，就3步， 和把大象放进冰箱里一样："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"numberedlist","attrs":{"start":null,"normalizeStart":1},"content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"打开冰箱门， 看看有没有空：用malloc申请空间。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"把大象装进冰箱里：strcpy把字符串拷贝到空间里。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":3,"align":null,"origin":null},"content":[{"type":"text","text":"关上冰箱门：不用的时候， free还回内存 （严谨的说，这里应该是先把大象请出来， 腾出冰箱的空间，以备下一次能够再装大象）。"}]}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"是不是很简单？需要的时候malloc申请内存，用完之后free释放内存。但实际上就这么简单的3行代码，可能会引发不少问题， 让我们step by step的看一下："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"问题1：如果上面第0步也变成 aMem = (char*) malloc(sizeof(char)), 这里直接执行line 1， 有什么问题？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"答: 内存泄漏，所有malloc申请的内存，必须要free释放之后才能再次被分配使用， 如果不free，那么程序会一直占用这段内存，直到整个进程结束。虽然程序逻辑执行没有问题， 但是如果内存泄漏过多，很可能在后面的程序中出现内存不足的问题，产生各种未知错误。但是要注意的是，如果第0步用malloc分配了空间给aMem，（假设地址是aMem=0x1234）,第1步这里的malloc同样分配了空间给aMem，（假设这次malloc返回地址是aMem=0x5678）, 也就是说， 0x1234指向的那段空间一直被占用，然后你的程序里却无法通过有效手段获得这个地址，也就没有办法再free它了。（因为aMem被修改成0x5678了）所以除非程序退出，不然我们再也没有机会释放这个0x1234指向的空间了。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"问题2：这里实际上申请了1024个byte的空间， 如果系统没有这么多空闲空间，有什么问题？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"答：直接报错， 这个时候要调查一下是不是存在内存泄漏。"},{"type":"text","marks":[{"type":"strong"}],"text":"问题3：如果copy的字符串不是“I am a bunch of memory”, 而是“1，2，3，4 ... 1025\" 会怎么样？"},{"type":"text","text":"答：由于strcpy不进行越界检查，这里第一步malloc出来的1024个字符， 却装载了1026个字符（包括'\\0'）， 也就是说内存被污染了， 这种情况轻的会导致内存溢出，如果被别有用心的人利用了， 可能就把你的程序所有信息dump出来...比如你的小秘密..."},{"type":"text","marks":[{"type":"strong"}],"text":"问题 4：如果之前内存没有申请成功，第3步free会有什么问题？"},{"type":"text","text":"答：出错，如果malloc之前失败了，其实就是第二步出错了。假设没有第二步， malloc失败之后，调用free程序会直接crash。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"问题 5：如果这里调用两次free会怎么样？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"答： 同样会出错， 两次free会导致未知错误、或程序crash。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"问题 6：如果这里free之后， aMem里面存的是什么值？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"答：free不会修改aMem的值，如果malloc之前返回0x1234给aMem，那么这里free之后，aMEM还是0x1234。试想一下，如果后面还用aMem访问0x1234会有什么问题？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"GC的意义"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"有人可能会说：上面6个问题完全可以避免， 只要我能保证malloc和free用的对就行啦。如果现实真的这么美好，那就万事大吉了。可惜现实情况是更为复杂的程序， 比如1000行的代码里存在if...else...、for \/while循环就会容易出现上面的问题。而且内存泄漏通常埋伏在你不知道的地方，慢慢积累，直到有一天产品的业务量达到一定程度后，服务进程就会突然崩溃。更可怕的是我们往往缺少有效的分析手段（或者高级的在线调试手段）来定位内存到底在哪里泄露了。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"所以除了严格执行编程规范，还有别的办法可以减少Memory leak吗？一些大牛们想到了一个办法：程序员只负责分配和使用内存，由计算机负责识别需要free释放的内存，并且自动把这些不用的内存free掉。这样程序员只要malloc\/new，不需要free\/delete。 如果计算机能识别并且回收不用的内存（垃圾），那么一方面减少了代码量，另一方面也会避免内存泄漏的可能性，岂不美哉？这就是Automatic Memory Management概念的由来，也就是GC的由来。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"现在大家应该明白GC的意义了吧，主要包括下面两方面："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"一方面减少开发者的代码成本。开发者无需关心内存如何回收，可以减少思考程序内存使用逻辑的时间。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"另一方面保证程序的正确性。没有了开发者的介入，减少了各种人为产生的内存泄漏和误free等问题，计算机更可以保证程序的正确性。程序就会更健壮， 也减少了运维人员半夜爬起来排障的机会。"}]}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"GC算法"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"下面我们来介绍GC里面各种牛闪闪的算法：Reference Counting，Mark Sweep，Concurrent Mark Sweep，Generational Concurrent Mark Sweep等，这些算法其实可粗略的分为两大类："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"一种是找到垃圾，回收之。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":"br"}}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"另一种是找到不是垃圾的对象，保留之。剩下的就作为垃圾对象，将它们回收掉。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":"br"}}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"由此而来，目前GC算法主要分为两类：Reference Counting(引用计数) 与 Object Tracing (对象追踪)。今天我们主要谈谈Reference Counting。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"Reference Counting"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"引用计数（Reference Counting）就是一种发现垃圾对象，并回收的算法。广义上讲，垃圾对象是指不再被程序访问的Object，具体细分的话，“不再被程序访问的对象”实际上还要分成两类。来来来，让我们对Object进行一次灵魂拷问：你是什么样的垃圾？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/9f\/a0\/9f5d216c3e31a94ff7197cf1e9ca32a0.png","alt":null,"title":"","style":[{"key":"width","value":"100%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"不再被程序访问的Object，具体可以细分为两大类："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#0052D9","name":"user"}},{"type":"strong"}],"text":"1. 对象被还能被访问到， 但是程序以后不再用到它了。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"举个例子："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"public class A {\nprivate void method() {\n System.out.println(\"I am a method\");\n}\npublic static void main (String args[]) {\n A a1 = new A();\n A a2 = new A();\n a1.method();\n \/\/ The following code has noting to do with a2\n ....\n .... \/\/ a2.method();\n}\n}"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"这个例子里面，a2还能被访问到，但是程序后面也不会用到它了。从程序逻辑角度，这个a2指向的对象就是垃圾，但是从计算机的角度，这个垃圾“不够垃圾”。因为如果程序后面突然后悔了，想用a2这个对象了 (比如code里面最后一行注释)， 程序还是可以正常访问到这个对象的。 所以从计算机的角度，a2所指向的对象不是垃圾。看到这里，大家可能会疑问：编码时已经注释了a2.method()，那么程序肯定不会运行这段代码， 这样的话，a2引用的对象还是垃圾，为什么从计算机的角度来讲a2对象却不是垃圾？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"实际上，我们有很多语言是支持动态代码修改的，比如Java的Bytecode Instrument，完全可以在运行时插入a2.method()的字节码，所以还是可以访问的。另外，这段代码的逻辑就是a2在函数栈上，a2引用的对象在堆里，所以只要a2一直引用这个对象，这个对象对程序来说可见的，计算机不会认为它是垃圾，所以这种垃圾是不可回收物。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"italic"},{"type":"strong"}],"text":"计算机： 我不要你觉得，我要我觉得！"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#0052D9","name":"user"}},{"type":"strong"}],"text":"2. 对象已经不能被访问了， 程序想用也没有办法找到它。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"还是举个例子："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"public class A {\nprivate void method() {\n System.out.println(\"I am a method\");\n}\npublic static void main (String args[]) {\n A a1 = new A();\n A a2 = new A();\n a1.method();\n \/\/ The following code has noting to do with a2\n ....\n ....\n a2 = a1;\n}\n}"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"和前面例子几乎一致，只是最后我们把a1赋值给a2。这里a2的值就变了，也就是说a2指向的对象变成了a1指向的对象，a2原来的对象就没有别的东西引用它了，程序在此之后没有任何办法可以访问到它。所以它就变成了真正的垃圾。请看下图："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/6b\/8e\/6b9052115442c2b7f4185f6a534f138e.png","alt":null,"title":"","style":[{"key":"width","value":"100%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"所以我们通常所讲的垃圾回收技术，也主要用来处理这种对象。那么问题来了， 如何找到这种对象呢？ 按照刚才的思路，没有再被任何东西引用的对象，就是可回收垃圾，由此得出一个简单直观的回收算法：引用计数。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"引用计数的概念， Wikipedia的解释："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"italic"}],"text":"In computer science, reference counting is a programming technique of storing the number of references, pointers, or handles to a resource, such as an object, a block of memory, disk space, and others."}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"简单说来就是以下几点："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"numberedlist","attrs":{"start":null,"normalizeStart":1},"content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"所有对象都存在一个记录引用计数的计数器，可能在对象里面，也可能单独的数据结构，总之是一种记录数据的地方。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"所有对象在创建的时候（比如new）, 引用计数为1。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":3,"align":null,"origin":null},"content":[{"type":"text","text":"当有别的变量或者对象对其进行引用，引用计数+1。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":4,"align":null,"origin":null},"content":[{"type":"text","text":"当有别的对象进行引用变更时，原先被引用的对象引用计数-1。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":5,"align":null,"origin":null},"content":[{"type":"text","text":"当引用计数为0的时候，回收对象"}]}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"看不懂？没关系，上代码："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"public class A {\nprivate void method() {\n System.out.println(\"I am a method\");\n}\npublic static void main (String args[]) {\n \/\/ 假设每个对象有一个引用计数变量rc\n A a1 = new A(); \/\/ 在堆上创建对象A, A.rc++;\n A a2 = new A(); \/\/ 在堆上创建对象A1,A1.rc++;\n a2 = a1; \/\/ A1.rc--,if ( A1.rc == 0 ) { 回收A1 }, A.rc++;\n} \/\/ 函数退出:\n \/\/ a1销毁, A.rc--;\n \/\/ a2销毁, A.rc--;\n \/\/ if ( A.rc == 0 ) { 回收A }\n}"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"还没看懂？上图："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/e2\/24\/e204c5870d801ce6e3c5355538543524.png","alt":null,"title":"","style":[{"key":"width","value":"100%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"读到这里，你应该就明白Reference Counting的核心原理了。看起来很简单，只需要一个计数器和一些加减法就可以进行内存回收了。但是，Reference Counting存在一个比较大的问题，也是我个人认为目前Reference Counting算法研究的核心问题：循环引用 。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"循环引用"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"请看下面的伪代码："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"class Parent {\n Child child;\n}\nclass Child {\n Parent parent;\n}\npublic class Main {\n public static void main (String[] args) {\n Parent p = new Parent();\n Child c = new Child();\n p.child = c;\n c.parent = p;\n }\n}"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"图就是这样的："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/a4\/02\/a48b58a05ec732f00365573343264102.png","alt":null,"title":"","style":[{"key":"width","value":"100%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"这个互相引用产生了环状引用， 引用计数器一致保持在1， Object无法被回收，造成了内存泄漏。可能你会问：不就是一个环，两个Object吗？这一点泄漏不是大问题，谁写代码不泄漏点内存。但是遇到下面这种情况呢？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/e9\/21\/e9f63a9738c22001a4ae70f1cb68a921.jpg","alt":null,"title":"","style":[{"key":"width","value":"100%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"单单一个环，带了一个长长的小尾巴，导致整个链上的所有对象无法回收，Heap内存逐渐失控，最终出现OOM异常，系统崩，代码卒。那么如何处理这个循环引用的问题呢？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"破环之道"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"就如前面所说， Reference Counting目前主要的研究课题都在破坏环形引用上。在我看来，目前主要是以下两种模式："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#0052D9","name":"user"}},{"type":"strong"}],"text":"1. 左边跟我一起画条龙: 把问题抛给程序员"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"就是在程序设计语言层面提供一些办法，可以是API、注解、新的关键字等等，然后把破环的能力交给程序员。比如Swift 提供的weak\/unbound关键字，包括C++的weak_ptr，相对于strong或者默认的引用，weak在进行引用时不做引用计数的增减，而是判断所引用的对象是否已经被回收，这样所有构成环的引用都用weak来做引用，这样在计数器中，构成环的部分就不计数了。这样做的优缺点是："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"优点：计算机不需要考虑环状问题，只要按照计数器进行对象回收就可以了。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"缺点：程序员的意识直接决定了内存会不会溢出。如果程序员不使用weak关键字，那么有可能造成上述的内存泄漏。"}]}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#0052D9","name":"user"}},{"type":"strong"}],"text":"2. 右边再划一道彩虹：把问题抛给计算机"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"这种办法就是让计算机自己找到方法去检测循环引用，一种常见的方法是配合Tracing GC，找到没有被环以外的对象引用的环，把它们回收掉。关于Tracing GC 咱们放到后续讨论。大家这里只要理解，为了帮助引用计数处理环形引用，计算机必须在适当的时候触发一个单独算法来找到环，然后再做处理。这样做的优缺点："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"优点：程序员完全不需要介入，只需专注自己的业务实现。weak pointer、strong pointer分不清楚也无所谓。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"缺点：需要加入新的环形引用检测机制，算法复杂度，对于程序的影响都是问题。"}]}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"说了这么多，咱们总结一下Reference Counting的优缺点："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#000000","name":"user"}},{"type":"strong"}],"text":"优点"},{"type":"text","marks":[{"type":"color","attrs":{"color":"#000000","name":"user"}}],"text":"："},{"type":"text","text":"Reference Counting算法设计简单，只需要在引用发生变化时进行计数就可以决定Object是否变成垃圾。并且可以随着对象的引用计数归零做到实时回收对象。所以Reference Counting是没有单独的GC阶段的，程序不会出现所谓的GC stop the world 阶段。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#000000","name":"user"}},{"type":"strong"}],"text":"缺点："},{"type":"text","text":"程序在运行过程中会不断的生成对象，给对象成员变量赋值，改变对象变量等等。这所有的操作都需要引入一次++和--，程序性能必然受影响。（目前一种优化方法就是利用编译器优化技术，减少Reference Counting引入的计数问题，但也无法完全避免）。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"处理环形引用问题，不论是交给程序员处理，还是交给计算机处理，都增加了程序的复杂度，还有可能会引入GC stop the world phase，这些都会在一定程度上影响程序的性能和吞吐量。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"好啦！今天就聊到这里吧，预知后事如何，且听下回分解！下次给大家分享另一类GC算法：Tracing GC，这也是目前应用比较广泛的一类算法。不论是Javascript的V8、Android的ART、Java的Hotspot、OpenJ9，还是Golang的GC，都采用了Tracing GC算法。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"horizontalrule"},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"头图：Unsplash"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"作者：臧琳"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"原文："},{"type":"link","attrs":{"href":"https:\/\/mp.weixin.qq.com\/s\/Qhu-Cpup4fobah6hsnGKpQ","title":null,"type":null},"content":[{"type":"text","text":"https:\/\/mp.weixin.qq.com\/s\/Qhu-Cpup4fobah6hsnGKpQ"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"原文: 趣谈GC技术，解密垃圾回收的玄学理论（一）"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"来源：腾讯云中间件 - 微信公众号 [ID：gh_6ea1bc2dd5fd]"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"转载：著作权归作者所有。商业转载请联系作者获得授权，非商业转载请注明出处。"}]}]}
趣谈GC技术，解密垃圾回收的玄学理论（一）
{"type":"doc","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"大多数程序员在日常开发中常常会碰到GC的问题：OOM异常、GC停顿等，这些异常直接导致糟糕的用户体验，如果不能得到及时处理，还会严重影响应用程序的性能。本系列从GC的基础入手，逐步帮助读者熟悉GC各种技术及问题根源。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"GC的由来"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"想当初，盘古开天辟地......"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"好吧，扯远了，这也不是仙侠小说..."}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"GC到底是怎么来的呢？这个问题要从C语言聊起， 大家都知道， C\/C++语言在编写程序的时候， 需要码神们自己管理内存， 直观的说就是使用内存的时候要malloc，之后这段内存会一直保留给程序进程，直到程序显式的调用free才会得以释放。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"一个例子引发的问题"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"\/\/第 0 步: char* aMem;\n\/\/第 1 步:\naMem = (char*) malloc(sizeof(char) * 1024);\n\/\/第 2 步:\nstrcpy(aMem, \"I am a bunch of memory\");\n\/\/第 3 步:\nfree(aMem);"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"看到没有，就3步， 和把大象放进冰箱里一样："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"numberedlist","attrs":{"start":null,"normalizeStart":1},"content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"打开冰箱门， 看看有没有空：用malloc申请空间。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"把大象装进冰箱里：strcpy把字符串拷贝到空间里。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":3,"align":null,"origin":null},"content":[{"type":"text","text":"关上冰箱门：不用的时候， free还回内存 （严谨的说，这里应该是先把大象请出来， 腾出冰箱的空间，以备下一次能够再装大象）。"}]}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"是不是很简单？需要的时候malloc申请内存，用完之后free释放内存。但实际上就这么简单的3行代码，可能会引发不少问题， 让我们step by step的看一下："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"问题1：如果上面第0步也变成 aMem = (char*) malloc(sizeof(char)), 这里直接执行line 1， 有什么问题？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"答: 内存泄漏，所有malloc申请的内存，必须要free释放之后才能再次被分配使用， 如果不free，那么程序会一直占用这段内存，直到整个进程结束。虽然程序逻辑执行没有问题， 但是如果内存泄漏过多，很可能在后面的程序中出现内存不足的问题，产生各种未知错误。但是要注意的是，如果第0步用malloc分配了空间给aMem，（假设地址是aMem=0x1234）,第1步这里的malloc同样分配了空间给aMem，（假设这次malloc返回地址是aMem=0x5678）, 也就是说， 0x1234指向的那段空间一直被占用，然后你的程序里却无法通过有效手段获得这个地址，也就没有办法再free它了。（因为aMem被修改成0x5678了）所以除非程序退出，不然我们再也没有机会释放这个0x1234指向的空间了。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"问题2：这里实际上申请了1024个byte的空间， 如果系统没有这么多空闲空间，有什么问题？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"答：直接报错， 这个时候要调查一下是不是存在内存泄漏。"},{"type":"text","marks":[{"type":"strong"}],"text":"问题3：如果copy的字符串不是“I am a bunch of memory”, 而是“1，2，3，4 ... 1025\" 会怎么样？"},{"type":"text","text":"答：由于strcpy不进行越界检查，这里第一步malloc出来的1024个字符， 却装载了1026个字符（包括'\\0'）， 也就是说内存被污染了， 这种情况轻的会导致内存溢出，如果被别有用心的人利用了， 可能就把你的程序所有信息dump出来...比如你的小秘密..."},{"type":"text","marks":[{"type":"strong"}],"text":"问题 4：如果之前内存没有申请成功，第3步free会有什么问题？"},{"type":"text","text":"答：出错，如果malloc之前失败了，其实就是第二步出错了。假设没有第二步， malloc失败之后，调用free程序会直接crash。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"问题 5：如果这里调用两次free会怎么样？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"答： 同样会出错， 两次free会导致未知错误、或程序crash。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"问题 6：如果这里free之后， aMem里面存的是什么值？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"答：free不会修改aMem的值，如果malloc之前返回0x1234给aMem，那么这里free之后，aMEM还是0x1234。试想一下，如果后面还用aMem访问0x1234会有什么问题？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"GC的意义"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"有人可能会说：上面6个问题完全可以避免， 只要我能保证malloc和free用的对就行啦。如果现实真的这么美好，那就万事大吉了。可惜现实情况是更为复杂的程序， 比如1000行的代码里存在if...else...、for \/while循环就会容易出现上面的问题。而且内存泄漏通常埋伏在你不知道的地方，慢慢积累，直到有一天产品的业务量达到一定程度后，服务进程就会突然崩溃。更可怕的是我们往往缺少有效的分析手段（或者高级的在线调试手段）来定位内存到底在哪里泄露了。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"所以除了严格执行编程规范，还有别的办法可以减少Memory leak吗？一些大牛们想到了一个办法：程序员只负责分配和使用内存，由计算机负责识别需要free释放的内存，并且自动把这些不用的内存free掉。这样程序员只要malloc\/new，不需要free\/delete。 如果计算机能识别并且回收不用的内存（垃圾），那么一方面减少了代码量，另一方面也会避免内存泄漏的可能性，岂不美哉？这就是Automatic Memory Management概念的由来，也就是GC的由来。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"现在大家应该明白GC的意义了吧，主要包括下面两方面："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"一方面减少开发者的代码成本。开发者无需关心内存如何回收，可以减少思考程序内存使用逻辑的时间。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"另一方面保证程序的正确性。没有了开发者的介入，减少了各种人为产生的内存泄漏和误free等问题，计算机更可以保证程序的正确性。程序就会更健壮， 也减少了运维人员半夜爬起来排障的机会。"}]}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"GC算法"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"下面我们来介绍GC里面各种牛闪闪的算法：Reference Counting，Mark Sweep，Concurrent Mark Sweep，Generational Concurrent Mark Sweep等，这些算法其实可粗略的分为两大类："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"一种是找到垃圾，回收之。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":"br"}}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"另一种是找到不是垃圾的对象，保留之。剩下的就作为垃圾对象，将它们回收掉。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":"br"}}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"由此而来，目前GC算法主要分为两类：Reference Counting(引用计数) 与 Object Tracing (对象追踪)。今天我们主要谈谈Reference Counting。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"Reference Counting"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"引用计数（Reference Counting）就是一种发现垃圾对象，并回收的算法。广义上讲，垃圾对象是指不再被程序访问的Object，具体细分的话，“不再被程序访问的对象”实际上还要分成两类。来来来，让我们对Object进行一次灵魂拷问：你是什么样的垃圾？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/9f\/a0\/9f5d216c3e31a94ff7197cf1e9ca32a0.png","alt":null,"title":"","style":[{"key":"width","value":"100%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"不再被程序访问的Object，具体可以细分为两大类："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#0052D9","name":"user"}},{"type":"strong"}],"text":"1. 对象被还能被访问到， 但是程序以后不再用到它了。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"举个例子："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"public class A {\nprivate void method() {\n System.out.println(\"I am a method\");\n}\npublic static void main (String args[]) {\n A a1 = new A();\n A a2 = new A();\n a1.method();\n \/\/ The following code has noting to do with a2\n ....\n .... \/\/ a2.method();\n}\n}"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"这个例子里面，a2还能被访问到，但是程序后面也不会用到它了。从程序逻辑角度，这个a2指向的对象就是垃圾，但是从计算机的角度，这个垃圾“不够垃圾”。因为如果程序后面突然后悔了，想用a2这个对象了 (比如code里面最后一行注释)， 程序还是可以正常访问到这个对象的。 所以从计算机的角度，a2所指向的对象不是垃圾。看到这里，大家可能会疑问：编码时已经注释了a2.method()，那么程序肯定不会运行这段代码， 这样的话，a2引用的对象还是垃圾，为什么从计算机的角度来讲a2对象却不是垃圾？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"实际上，我们有很多语言是支持动态代码修改的，比如Java的Bytecode Instrument，完全可以在运行时插入a2.method()的字节码，所以还是可以访问的。另外，这段代码的逻辑就是a2在函数栈上，a2引用的对象在堆里，所以只要a2一直引用这个对象，这个对象对程序来说可见的，计算机不会认为它是垃圾，所以这种垃圾是不可回收物。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"italic"},{"type":"strong"}],"text":"计算机： 我不要你觉得，我要我觉得！"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#0052D9","name":"user"}},{"type":"strong"}],"text":"2. 对象已经不能被访问了， 程序想用也没有办法找到它。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"还是举个例子："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"public class A {\nprivate void method() {\n System.out.println(\"I am a method\");\n}\npublic static void main (String args[]) {\n A a1 = new A();\n A a2 = new A();\n a1.method();\n \/\/ The following code has noting to do with a2\n ....\n ....\n a2 = a1;\n}\n}"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"和前面例子几乎一致，只是最后我们把a1赋值给a2。这里a2的值就变了，也就是说a2指向的对象变成了a1指向的对象，a2原来的对象就没有别的东西引用它了，程序在此之后没有任何办法可以访问到它。所以它就变成了真正的垃圾。请看下图："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/6b\/8e\/6b9052115442c2b7f4185f6a534f138e.png","alt":null,"title":"","style":[{"key":"width","value":"100%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"所以我们通常所讲的垃圾回收技术，也主要用来处理这种对象。那么问题来了， 如何找到这种对象呢？ 按照刚才的思路，没有再被任何东西引用的对象，就是可回收垃圾，由此得出一个简单直观的回收算法：引用计数。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"引用计数的概念， Wikipedia的解释："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"italic"}],"text":"In computer science, reference counting is a programming technique of storing the number of references, pointers, or handles to a resource, such as an object, a block of memory, disk space, and others."}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"简单说来就是以下几点："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"numberedlist","attrs":{"start":null,"normalizeStart":1},"content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"所有对象都存在一个记录引用计数的计数器，可能在对象里面，也可能单独的数据结构，总之是一种记录数据的地方。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"所有对象在创建的时候（比如new）, 引用计数为1。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":3,"align":null,"origin":null},"content":[{"type":"text","text":"当有别的变量或者对象对其进行引用，引用计数+1。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":4,"align":null,"origin":null},"content":[{"type":"text","text":"当有别的对象进行引用变更时，原先被引用的对象引用计数-1。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":5,"align":null,"origin":null},"content":[{"type":"text","text":"当引用计数为0的时候，回收对象"}]}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"看不懂？没关系，上代码："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"public class A {\nprivate void method() {\n System.out.println(\"I am a method\");\n}\npublic static void main (String args[]) {\n \/\/ 假设每个对象有一个引用计数变量rc\n A a1 = new A(); \/\/ 在堆上创建对象A, A.rc++;\n A a2 = new A(); \/\/ 在堆上创建对象A1,A1.rc++;\n a2 = a1; \/\/ A1.rc--,if ( A1.rc == 0 ) { 回收A1 }, A.rc++;\n} \/\/ 函数退出:\n \/\/ a1销毁, A.rc--;\n \/\/ a2销毁, A.rc--;\n \/\/ if ( A.rc == 0 ) { 回收A }\n}"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"还没看懂？上图："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/e2\/24\/e204c5870d801ce6e3c5355538543524.png","alt":null,"title":"","style":[{"key":"width","value":"100%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"读到这里，你应该就明白Reference Counting的核心原理了。看起来很简单，只需要一个计数器和一些加减法就可以进行内存回收了。但是，Reference Counting存在一个比较大的问题，也是我个人认为目前Reference Counting算法研究的核心问题：循环引用 。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"循环引用"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"请看下面的伪代码："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"class Parent {\n Child child;\n}\nclass Child {\n Parent parent;\n}\npublic class Main {\n public static void main (String[] args) {\n Parent p = new Parent();\n Child c = new Child();\n p.child = c;\n c.parent = p;\n }\n}"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"图就是这样的："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/a4\/02\/a48b58a05ec732f00365573343264102.png","alt":null,"title":"","style":[{"key":"width","value":"100%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"这个互相引用产生了环状引用， 引用计数器一致保持在1， Object无法被回收，造成了内存泄漏。可能你会问：不就是一个环，两个Object吗？这一点泄漏不是大问题，谁写代码不泄漏点内存。但是遇到下面这种情况呢？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/e9\/21\/e9f63a9738c22001a4ae70f1cb68a921.jpg","alt":null,"title":"","style":[{"key":"width","value":"100%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"单单一个环，带了一个长长的小尾巴，导致整个链上的所有对象无法回收，Heap内存逐渐失控，最终出现OOM异常，系统崩，代码卒。那么如何处理这个循环引用的问题呢？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"破环之道"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"就如前面所说， Reference Counting目前主要的研究课题都在破坏环形引用上。在我看来，目前主要是以下两种模式："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#0052D9","name":"user"}},{"type":"strong"}],"text":"1. 左边跟我一起画条龙: 把问题抛给程序员"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"就是在程序设计语言层面提供一些办法，可以是API、注解、新的关键字等等，然后把破环的能力交给程序员。比如Swift 提供的weak\/unbound关键字，包括C++的weak_ptr，相对于strong或者默认的引用，weak在进行引用时不做引用计数的增减，而是判断所引用的对象是否已经被回收，这样所有构成环的引用都用weak来做引用，这样在计数器中，构成环的部分就不计数了。这样做的优缺点是："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"优点：计算机不需要考虑环状问题，只要按照计数器进行对象回收就可以了。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"缺点：程序员的意识直接决定了内存会不会溢出。如果程序员不使用weak关键字，那么有可能造成上述的内存泄漏。"}]}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#0052D9","name":"user"}},{"type":"strong"}],"text":"2. 右边再划一道彩虹：把问题抛给计算机"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"这种办法就是让计算机自己找到方法去检测循环引用，一种常见的方法是配合Tracing GC，找到没有被环以外的对象引用的环，把它们回收掉。关于Tracing GC 咱们放到后续讨论。大家这里只要理解，为了帮助引用计数处理环形引用，计算机必须在适当的时候触发一个单独算法来找到环，然后再做处理。这样做的优缺点："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"优点：程序员完全不需要介入，只需专注自己的业务实现。weak pointer、strong pointer分不清楚也无所谓。"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"缺点：需要加入新的环形引用检测机制，算法复杂度，对于程序的影响都是问题。"}]}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"说了这么多，咱们总结一下Reference Counting的优缺点："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#000000","name":"user"}},{"type":"strong"}],"text":"优点"},{"type":"text","marks":[{"type":"color","attrs":{"color":"#000000","name":"user"}}],"text":"："},{"type":"text","text":"Reference Counting算法设计简单，只需要在引用发生变化时进行计数就可以决定Object是否变成垃圾。并且可以随着对象的引用计数归零做到实时回收对象。所以Reference Counting是没有单独的GC阶段的，程序不会出现所谓的GC stop the world 阶段。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#000000","name":"user"}},{"type":"strong"}],"text":"缺点："},{"type":"text","text":"程序在运行过程中会不断的生成对象，给对象成员变量赋值，改变对象变量等等。这所有的操作都需要引入一次++和--，程序性能必然受影响。（目前一种优化方法就是利用编译器优化技术，减少Reference Counting引入的计数问题，但也无法完全避免）。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"处理环形引用问题，不论是交给程序员处理，还是交给计算机处理，都增加了程序的复杂度，还有可能会引入GC stop the world phase，这些都会在一定程度上影响程序的性能和吞吐量。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"好啦！今天就聊到这里吧，预知后事如何，且听下回分解！下次给大家分享另一类GC算法：Tracing GC，这也是目前应用比较广泛的一类算法。不论是Javascript的V8、Android的ART、Java的Hotspot、OpenJ9，还是Golang的GC，都采用了Tracing GC算法。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"horizontalrule"},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"头图：Unsplash"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"作者：臧琳"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"原文："},{"type":"link","attrs":{"href":"https:\/\/mp.weixin.qq.com\/s\/Qhu-Cpup4fobah6hsnGKpQ","title":null,"type":null},"content":[{"type":"text","text":"https:\/\/mp.weixin.qq.com\/s\/Qhu-Cpup4fobah6hsnGKpQ"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"原文: 趣谈GC技术，解密垃圾回收的玄学理论（一）"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"来源：腾讯云中间件 - 微信公众号 [ID：gh_6ea1bc2dd5fd]"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"转载：著作权归作者所有。商业转载请联系作者获得授权，非商业转载请注明出处。"}]}]}
發表評論
所有評論
還沒有人評論，想成為第一個評論的人麼? 請在上方評論欄輸入並且點擊發布.
相關文章