关于java:并发王者课青铜6借花献佛如何格式化Java内存工具JOL输出

5次阅读

共计 3312 个字符，预计需要花费 9 分钟才能阅读完成。

在后面的文章《一探到底 - 如何从 synchronized 了解 Java 对象头中的锁》中，咱们介绍并体验了 JOL 工具。尽管 JOL 很赞，但它的输入对咱们不是很敌对，如果不借助工具，咱们很难直观了解其中的含意。

上面这段代码是对 JOL 输入的翻译，倡议你珍藏。代码非我所写，文末曾经注明出处。


import org.openjdk.jol.info.ClassLayout;

import java.nio.ByteOrder;
import java.util.ArrayList;
import java.util.Collections;
import java.util.List;
import java.util.regex.Matcher;
import java.util.regex.Pattern;

public class PrintObjectHeader {
    /**
     * Get binary data
     *
     * @param o
     * @return
     */
    public static String getObjectHeader(Object o) {ByteOrder order = ByteOrder.nativeOrder();//Byte order
        String table = ClassLayout.parseInstance(o).toPrintable();
        Pattern p = Pattern.compile("(0|1){8}");
        Matcher matcher = p.matcher(table);
        List<String> header = new ArrayList<String>();
        while (matcher.find()) {header.add(matcher.group());
        }
        //Little-endian machines, need to traverse in reverse
        StringBuilder sb = new StringBuilder();
        if (order.equals(ByteOrder.LITTLE_ENDIAN)) {Collections.reverse(header);
        }
        for (String s : header) {sb.append(s).append(" ");
        }
        return sb.toString().trim();
    }

    /**
     * Parsing object header function for 64bit jvm
     * In 64bit jvm, the object header has two parts: Mark Word and Class Pointer, Mark Word takes 8 bytes, Class Pointer takes 4 bytes
     *
     * @param s Binary string of object header (each 8 bits, separated by a space)
     */
    public static void parseObjectHeader(String s) {String[] tmp = s.split(" ");
        System.out.print("Class Pointer:");
        for (int i = 0; i < 4; ++i) {System.out.print(tmp[i] + " ");
        }
        System.out.println("\nMark Word:");
        if (tmp[11].charAt(5) == '0' && tmp[11].substring(6).equals("01")) {//0 01 lock-free state, regardless of GC mark
            //notice: Mark word structure without lock: unused(25bit) + hashcode(31bit) + unused(1bit) + age(4bit) + biased_lock_flag(1bit) + lock_type(2bit)
            // The reason why hashcode only needs 31bit is: hashcode can only be greater than or equal to 0, eliminating the negative range, so you can use 31bit to store
            System.out.print("\thashcode (31bit):");
            System.out.print(tmp[7].substring(1) + " ");
            for (int i = 8; i < 11; ++i) System.out.print(tmp[i] + " ");
            System.out.println();} else if (tmp[11].charAt(5) == '1' && tmp[11].substring(6).equals("01")) {//1 01, which is the case of biased lock
            //notice: The object is in a biased lock, its structure is: ThreadID(54bit) + epoch(2bit) + unused(1bit) + age(4bit) + biased_lock_flag(1bit) + lock_type(2bit)
            // ThreadID here is the thread ID holding the biased lock, epoch: a timestamp of the biased lock, used for optimization of the biased lock
            System.out.print("\tThreadID(54bit):");
            for (int i = 4; i < 10; ++i) System.out.print(tmp[i] + " ");
            System.out.println(tmp[10].substring(0, 6));
            System.out.println("\tepoch:" + tmp[10].substring(6));
        } else {//In the case of lightweight locks or heavyweight locks, regardless of the GC mark
            //notice: JavaThread*(62bit,include zero padding) + lock_type(2bit)
            // At this point, JavaThread* points to the monitor of the lock record/heavyweight lock in the stack
            System.out.print("\tjavaThread*(62bit,include zero padding):");
            for (int i = 4; i < 11; ++i) System.out.print(tmp[i] + " ");
            System.out.println(tmp[11].substring(0, 6));
            System.out.println("\tLockFlag (2bit):" + tmp[11].substring(6));
            System.out.println();
            return;
        }
        System.out.println("\tage (4bit):" + tmp[11].substring(1, 5));
        System.out.println("\tbiasedLockFlag (1bit):" + tmp[11].charAt(5));
        System.out.println("\tLockFlag (2bit):" + tmp[11].substring(6));

        System.out.println();}

    public static void printObjectHeader(Object o) {if (o == null) {System.out.println("null object.");
            return;
        }
        parseObjectHeader(getObjectHeader(o));
    }
}

参考资料