Re: finitequeue heavy synchronization

Quartz wrote,
> It's kind'of important, because it could causes a lot of
> synchronization overhead for something as simple as going to memory
> instead of registers.

"could"? Have you got any benchmarks that show there's actually a 
performance issue here?

> So, which JVM are "volatile" compliant?

Strictly speaking, probably none. The original Java memory model 
entailed strict sequential consistency, and I believe that no 
mainstream JVMs actually implemented that policy (partly oversight, 
partly because it would prohibit many desirable optimizations at the 
JIT level). JSR-133 is changing the JMM to allow correct and efficient 
implementations, and 1.5 should support it.

That said, I'm pretty sure that Sun VMs post 1.4.1 implement volatile 
semantics correctly.

Cheers,


Miles


-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf