Concurrent programming -java memory model

1. Basic Concepts

　　Program: Static, the code used to complete certain functions.

　　Process: dynamic, running program

　　Thread: The actual operation of the unit process, a process can contain one or more threads.

2. JVM memory area

• Heap: threads share, store instance object (OOM)
• Virtual machine stack: the private address of the thread, Java methods at runtime memory model (OOM), the local variable storage, data reference type, the operand stack
• Native method stacks
• Methods District: threads share, store class information, constants, static variables, etc.
• Program Counter: thread private, storage address of the next instruction

3. java memory model (java memory model, JMM, abstract models)

Role: interactive specification memory space and work space data

Main memory: information sharing by threads

Working memory: thread private information. Basic data types, assigned directly to the working memory. Address reference stored in working memory, object references are stored in a heap.

Way of working:

　　Thread modify private data, direct change in the working space

　　Thread modifies shared data, copy data to a workspace, modify the workspace, the changes are complete, refresh the memory.

4. Hardware Memory Architecture

CPU cache coherency problem solution:

1. Bus lock: CPU throughput reduction

2. The cache coherence protocol on

When the CPU in the CACHE operation data, if the data is shared variable, in CACHE data read register, a new modified, and update the memory data

CaCHE LINE is set invalid, the other CPU can read data from memory

 5. java thread hardware processor

6. The three concurrent programming characteristics

Atomicity: x = 1 indivisible

Visibility: thread can manipulate the data in their own workspace

Ordering: the sequence program is not necessarily the order (re-compiled execution ordering, instruction reordering, Objective: efficiency

JMM three features ensure

1. JMM and Atomicity

1. X = 10 atomic write if it is a private data having atomic, if the shared data is not atomic (read-write)  
2. Y = x atomic no
1. X workspace data read (atomicity)
2. Writes the value of X Y (atomic)
3. No atomic I ++
1. I read into the workspace
2. +1；
3. Refresh results to memory
4. Z = z + 1 atomic no
1. Z read into the workspace
2. +1；
3. Refresh results to memory

A plurality of atomic operations to merge together without atomicity

Of guarantees:

Synchronized

JUC   Lock的lock

 

JMM与可见性

     Volatile

     Synchronized:加锁

     JUC   JUC   Lock的lock

 

1. JMM与有序性 

  Volatile：

  Synchronized：

Happens-before原则：

1. 程序次序原则
2. 锁定原则  ：后一次加锁必须等前一次解锁
3. Volatile原则：霸道原则
4. 传递原则：A---B ---C    A--C