JMM Memory model

order

since：2019 year 9 month 2 Japan  9:11
auth： Hadi
update： 2021 year 5 month 16 Japan  22:36
 Reference resources ：
	https://blog.csdn.net/lxm55913153/article/details/79208126

JMM

Java Memory model Java Memort Model Referred to as JMM.
JMM Defined java How virtual machines work in memory , It is a logical structure .
JMM It defines the visibility of shared variables between multiple threads and synchronizes shared variables when necessary .

JMM Operation type

JMM In order to share data , The operations on variables are divided into the following 8 class ：

• read
  Read data from main memory .
• load
  Write the data in the main memory into the working memory （ A child thread ）
• use
  Read data from working memory to calculate
• assign
  Re write the calculated value into the working memory
• store
  Write the data of working memory back to main memory
• write
  take store The written variable is assigned to the variable location in the main memory
• lock
  Lock the main memory variables , Identify that the variable is in thread exclusive state
• unlock
  Unlock the main memory variable , After unlocking, other threads can lock the variable

Please refer to the following flow chart for the specific process ：

PS： The operations in the red box are also carried out in the main memory .
Main memory main memory, Local memory local memory
local memory It doesn't really exist , Just an abstract concept , Covering caching , Write buffer , Registers and other hardware and compiler optimizations .

Competition and bus locking

Competition

If multiple threads share an object , And multiple threads modify this object at the same time , Then there will be competition . Here's the picture ：

Two threads are working on a variable at the same time ++ The operation of , Then it may cause the retrieval to be 1, When it is written back to the main memory, it is written as 1, Cause loss once add operation .

Bus yoke

For the occurrence of the above situation , We can operate lock The operation of , It will lead to low performance , Not efficient .

Cpu Read data from main memory to cache , This data will be locked on the bus ,
So other Cpu Can't read or write this data , Until this Cpu After using the data lock, others Cpu To read the data

MESI Cache consistency protocol

Multiple Cpu Read the same data from main memory to their respective cache , When one of them Cpu Modified the data in the cache , The data will be synchronized back to main memory immediately , Other Cpu Through the bus sniffing mechanism, we can sense the change of data and invalidate the data in our own cache , And data re read、load.

Volatile How cache visibility is implemented

The underlying implementation is mainly through assembly lock Prefix instruction , It locks the cache of this memory area （ Cache row locking ） And write back to main memory .
stay IA-32 Architecture software developer's Manual for lock Interpretation of instructions ：

• The data of the current processor cache line will be written back to the system memory immediately .
• This write back operation will cause other CPU Invalid data cached in memory address （MESI agreement ）
  

Java Program assembly code view

-server -Xcomp -XX:+UnlockDiagnosticVMOptions -XX:+PrintAssembly -XX:CompileCommand=compileonly,*VolatileVisibilityTest,prepareData