A detailed introduction to JVM bytecode in Java

This article brings you a detailed introduction to JVM bytecode in Java. It has certain reference value. Friends in need can refer to it. I hope it will be helpful to you.

This is an article about Java Basics (JVM). I originally wanted to talk about the Java class loading mechanism first, but then I thought about it. The role of the JVM is to load the bytecode compiled by the compiler and interpret it into the machine. code, then you should first understand the bytecode, and then talk about the class loading mechanism for loading the bytecode. It seems better, so this article is changed to a detailed explanation of the bytecode.

Due to the purely object-oriented nature of Java, as long as the bytecode can represent the information of a class, it can represent the entire Java program. As long as the JVM can load the information of a class, it can load the entire program. Therefore, whether it is bytecode or JVM loading mechanism, the focus is on classes. My main concerns are:

1. Since the bytecode is not loaded into memory all at once, how does the JVM know where the class information it wants to load is located in the .class file?

2. How does bytecode represent class information?

3. Will the bytecode optimize the program?

The first question is very simple, because even if a source file has many classes (only one public class), the compiler will generate a .class file for each class, and the JVM will load it as needed Just load the loaded class name.

To solve the following problems, first let's look at the composition of bytecode (open with Hex Fiend on Mac).

For this piece of code:

package com.test.main1;

public class ByteCodeTest {

	int num1 = 1;
	int num2 = 2;

	public int getAdd() {

		return num1 + num2;
	}
}

class Extend extends ByteCodeTest {
	public int getSubstract() {
		return num1 - num2;
	}
}

Let’s analyze the Extend class in it.

Use Hex Fiend to open the compiled .class file like this (hexadecimal code):

Since the class file has no delimiter, each What each position represents, the length of each part and other formats are strictly regulated, see the table below:

among them u1, u2, u4, u8 Represents an unsigned number of several bytes. In the decompiled hexadecimal file, two numbers represent one byte, which is u1.

Look at it one by one from beginning to end:

(1) magic: u4, the magic number, means that this file is a .class file. .jpg, etc. will also have this magic number. Because of the magic number, even if *.jpg is changed to *.123, it can still be opened as usual.

(2) minor version, major version: each u2, version number, backward compatible, that is, higher version JDK can use lower version .class files, but not vice versa.

(3) constant_pool_count: u2, the number of constants in the constant pool, 0019 represents 24.

(4) Next are the specific constants, a total of constant_pool_count-1.

Constant pool usually stores two types of data:

Literals: such as strings, final modified constants, etc.;

Symbol references: such as the full name of a class/interface Qualified names, method names and descriptions, field names and descriptions, etc.

According to the decompiled numbers, first check the table below to get the type and length of the constant. The next number that is equal to the length is the specific value of the constant.

For example, 070002, it means that the type is CONSTANT_Class_info, its tag is u1, and the length of u2 is the index pointing to the fully qualified name constant item. This index should also be viewed together with the class file opened by javap -verbose. The contents and order in the constant pool are clearly listed here:

You can see 0002 here The constant of the index item is: com/test/main1/Extend, which is the fully qualified name of the class. If the value is a string, you need to convert the value into decimal and check the ASCII code table to get the specific characters. The following constants are analyzed as follows:

01001563 6F6D2F74 6573742F 6D61696E 312F4578 74656E64：com/test/main1/Extend

070004：com/test/main1/ByteCodeTest

01001B63 6F6D2F74 6573742F 6D61696E 312F4279 7465436F 64655465 7374：com/test/main1/ByteCodeTest

0100063C 696E6974 3E：

01000328 2956: ()V

01000443 6F6465 ：Code

0A000300 09：com/test/main1/ByteCodeTest、"":()V

0C000500 06：、()V

01000F4C 696E654E 756D6265 72546162 6C65：LineNumberTable

0100124C 6F63616C 56617269 61626C65 5461626C 65：LocalVariableTable

01000474 6 86973：this

0100174C 636F6D2F 74657374 2F6D6169 6E312F45 7874656E 643B：Lcom/test /main1/Extend;

01000C67 65745375 62737472 616374：getSubstract

01000328 2949: ()I

09000100 11：com/test/main1/Extend、num1:I

0C001200 13: num1, I

0100046E 756D31: num1

01000149: I

09000100 15: com/test/main1/Extend, num2:I

0C001600 13: num2, I

0100046E 756D32: num2

01000A53 6F757263 6546696C 65: SourceFile

01001142 79746543 6F646554 6573742E 6A617661: ByteCodeTest.java

At this point, all the constants in the constant pool have been parsed.

(5) Next is the access_flags of u2: The main purpose of the access_flags access flag is to mark whether the class is a class or an interface. If it is a class, whether the access permission is public, whether it is abstract, and whether it is marked as final, etc., see the table below:

filed_info: {
　　u2　　　　　　　　access_flags;　　　
　　u2　　　　　　　　name_index;
　　u2　　　　　　　　descriptor_index;　　　　　　
　　u2　　　　　　　　attributes_count;
　　attribute_info　　　attributes[attributes_count];
}

method_info {
　　u2　　　　　　　　access_flags;　　　　　　　　　　0000　　　　？
　　u2　　　　　　　　name_index;　　　　　　　　　　 0005　　　<init>
　　u2　　　　　　　　descriptor_index;　　　　　　　　 0006　　　　()V
　　u2　　　　　　　　attributes_count;　　　　　　　　 0001　　　　1个
　　attribute_info　　　attributes[attributes_count];　　　0007　　　　Code
}

attribute_info {
　　u2　　attribute_name_index;　　0007　　Code
　　u1　　attribute_length;
　　u1　　info[attribute_length];
}

Code_attribute {　　//Code_attribute包含某个方法、实例初始化方法、类或接口初始化方法的Java虚拟机指令及相关辅助信息
　　u2　　attribute_name_index;　　0007　　　　　　　Code
　　u4　　attribute_length;　　　　0000002F　　　　　 47
　　u2　　max_stack;　　　　　　 0001　　　　　　　　1　//用来给出当前方法的操作数栈在方法执行的任何时间点的最大深度
　　u2　　max_locals;　　　　　　0001　　　　　 　　　1　//用来给出分配在当前方法引用的局部变量表中的局部变量个数
　　u4　　code_length;　　　　　 00000005　　　　　　5　//给出当前方法code[]数组的字节数
　　u1　　code[code_length];　　  2AB70008 B1　　　　42、183、0、8、177　
　　　　　　　　//给出了实现当前方法的Java虚拟机代码的实际字节内容　（这些数字代码实际对应一些Java虚拟机的指令）
　　u2　　exception_table_lentgh; 　　0000　　　　　　 0　　　　　　　  　//异常的信息
　　{
　　　　u2　　start_pc;　　　//这两项的值表明了异常处理器在code[]中的有效范围，即异常处理器x应满足：start_pc≤x≤end_pc
　　　　u2　　end_pc;　　　//start_pc必须在code[]中取值，end_pc要么在code[]中取值，要么等于code_length的值
　　　　u2　　handler_pc;　//表示一个异常处理器的起点
　　　　u2　　catch_type;　//表示当前异常处理器需要捕捉的异常类型。为0，则都调用该异常处理器，可用来实现finally。
　　} exception_table[exception_table_lentgh];　　　　　　在本类中大括号里的结构为空
　　u2　　attribute_count;　　　　0002　　　　　　　　　　2　　　　表示该方法的其它附加属性，本类有1个
　　attribute_info　　attributes[attributes_count];　　　　　  000A、000B　　　  LineNumberTable、LocalVariableTable
}

LineNumberTable_attribute {　//被调试器用来确定源文件中由给定的行号所表示的内容，对应于Java虚拟机code[]数组的哪部分
　　u2　　attribute_name_index;　　　　　　000A
　　u4　　attribute_length;　　　　 　　　　 00000006
　　u2　　line_number_table_length;　　　　 0001
　　{　　u2　　start_pc;　　　　　　　　　　0000　　　　
　　　　 u2　　line_number;　　　　　　　　 000E　　　　　//该值必须与源文件中对应的行号相匹配
　　} line_number_table[line_number_table_length];
}

LocalVariableTable_attribute {
　　u2　　attribute_name_index;　　　　000B
　　u4　　attribute_length;　　　　　　  0000000C
　　u2　　local_variable_table_length; 　 0001
　　{　　u2　　start_pc;　　　　　　　   0000
　　　　 u2　　length;　　　　　　　　   0005
　　　　 u2　　name_index;　　　　　　 000C
　　　　 u2　　descriptor_index;　　　　000D　　　　//用来表示源程序中局部变量类型的字段描述符
　　　　u2　　 index;　　　　　　　　　 0000
　　} local_variable_table[local_variable_table_length];

SourceFile_attribute {
　　u2　　attribute_name_index;　　0017　　　　　SourceFile
　　u4　　attribute_length;　　　　  00000002　　 2
　　u2　　sourcefile_index;　　　     0018　　　　　ByteCodeTest.java　//表示本class文件是由ByteCodeTest.java编译来的
}