JVM memory model and local memory

Java Class instances of are generally in  JVM  Heap allocation , and Java It's through JNI call C Code to implement Socket communication , that C Where is the memory required by the code to be run allocated ？C Whether the code can operate directly Java Pile up ？

To answer these questions , Let me start with JVM Relationship with user processes . If you want to run a Java Class file , You can use the following Java Command to execute

java my.class

This Command line Medium java It is actually an executable program , This program will create JVM To load and run your Java class .

The operating system will create a process to execute this java Executable program , Each process has its own virtual address space ,JVM Use of Memory （ Including heaps 、 Stack and method area ） It is allocated from the virtual address space of the process . Please note that ,JVM Memory is just a part of the process space , In addition, there are code segments in the process space 、 Data segment 、 Memory mapped area 、 Kernel space .JVM From the perspective of ,JVM The part outside of memory is called local memory ,C The memory used by the program code in the running process is allocated in the local memory . Let's use a picture to understand .

There is something wrong with this diagram , The method area is already in Java8 Time is put into local memory , Named meta space ——> Deposit class Class loaded class information

that HeapByteBuffer and DirectByteBuffer What's the difference ？HeapByteBuffer The object itself is JVM Heap allocation , And it holds an array of bytes byte[] Also in JVM Heap allocation .

But if you use HeapByteBuffer To receive network data , You need to copy the data from the kernel to a temporary local memory , Copy from temporary local memory to JVM Pile up , Instead of copying directly from the kernel to JVM Pile it up . Why is that ？ This is because the data is copied from the kernel to JVM In the process of stacking ,JVM It could happen GC,GC Objects may be moved during the process , in other words JVM The byte array on the heap may be moved , In this case Buffer The address is invalid . If this is transferred through local memory , From local memory to JVM During the copy of the heap JVM Can guarantee not to do GC.

If you use HeapByteBuffer, You'll find that JVM There is a layer of transition between the heap and the kernel , and DirectByteBuffer To solve this problem ,DirectByteBuffer The object itself is JVM Pile it up , But the byte array it holds is not from JVM Allocated on the heap , It is allocated from local memory .
DirectByteBuffer One of the objects is long The type field address, The local address of the memory is recorded , So when receiving data , Pass this local memory address directly to C Program ,C The program will copy the network data from the kernel to this local memory ,JVM You can read this local memory directly , This way ratio HeapByteBuffer One less copy , So generally speaking, it will be faster than HeapByteBuffer Several times faster . You can deepen your understanding through the above figure .

What then? HeapByteBuffer Performance ratio of DirectByteBuffer The difference is still in use ？
This is because the local memory is difficult to manage , It is difficult to locate a memory leak , In terms of stability ,
HeapByteBuffer Better ;

relevant ：

(26 Bar message ) JVM- How bytecode is JVM Executive + A little thought about threads _Fairy want carry The blog of -CSDN Blog

(26 Bar message ) Garbage collection process - A brief account _Fairy want carry The blog of -CSDN Blog _ Garbage collection process

(26 Bar message ) Class loading some understanding （—）_Fairy want carry The blog of -CSDN Blog