What are the differences between the volatile in c# and java?

Question

In some .net documentation I have read this is how the compiler handles volatile:

• *"Reading from a volatile or using the Thread.VolatileRead method is logically an acquire fence"
• "Writing to a volatile or using the Thread.VolatileWrite method is logically a release fence" *

These fences apply at both complier and architecture level.

Of course the main difference in VC++ is that the fence is only applied at complier level.

So my question is, what are the memory reordering prevention semantics of volatile in Java?

Conversion:

Fence = Barrier Barrier = Fence

References:

Joe Duffy (Concurrent Programming on Windows)

Answer 1

In Java, the javac compiler does next to nothing with volatile. It doesn't re-order statements and does almost no optimisation.

The JIT on the other hand can do quite a bit of optimisation and re-ordering.

The important features of volatile are;

• read/write access cannot be optimise away
• any write which occurs before a volatile write has to occur before the write.
• any read which occurs after a volatile read, must occur after the read.