[concurrent programming 053] Why use volatile variables for double check lock variables

Double check why lock variables are used volatile Variable

Look at the notes in the figure , Assuming that thread A perform getInstance Method

Up to the third day 4 Step ： instance by null, entering if Judge ;

Up to the third day 5 Step ： obtain synchronized lock , success , Go to sync code block ;

Up to the third day 6 Step ： Continue to judge instance, by null entering if Judge ;

Up to the third day 7 Step ： instance = new Instance() . It seems to be a sentence of code , It's actually three sentences of code .

Up to the third day 7.1 Step ： Allocate memory space for objects .

Execute to 7.3 Step ： about 7.2 Step sum 7.3 Step , There is no dependency between the two , Set up instance Points to the memory address just allocated .

Just in the thread A After execution 7.3 After step , Not yet implemented 7.2 When initializing objects , Threads B coming .

Look at the threads B：

Execution section 4 Step ： instance Not for null, Do not enter if Judge ; （ Why at this time instance Not for null？ because == The judgment of the , The judgment is internal Storage address , Threads A Although the object is not initialized , But it has been set instance Point to the memory address , So mistakenly think that at this time instance No by null）

Execution section 10 Step ： return instance; （ Because the initialization object has not been executed yet , here instance The corresponding instance object is null） .

Something is wrong. .

So join volatile, prohibit 7.2 and 7.3 The reordering of , Problem solved .