On Wed, Aug 12, 2009 at 4:24 PM, Richard Newman<holyg...@gmail.com> wrote:
>
>> I may be wrong, but doesn't a typical function invocation involve
>> dereferencing the Var holding the object that implements "IFn" and
>> calling invoke? It seems pretty intuitive to me that this would be
>> difficult to inline by the JIT, there is a little bit of
>> synchronization going on every time a Var is dereferenced.
>
> In principle, the JIT can inline the Var lookup, and do the
> appropriate analysis to eliminate much of the work -- Vars have thread-
> local bindings, so the JVM should be pretty well aware of access and
> scope. Of course, this will only happen if everything is small enough,
> frequently used, etc. etc.
>
> I saw a presentation at JavaOne which illustrated to just what extent
> the dynamic compiler can eliminate locks, allocations, aliases,
> synchronization boundaries, do closed-world analysis of class
> hierarchies, and so on. It's pretty impressive. ("Inside Out: A Modern
> Virtual Machine Revealed", if you're interested.)
>
I don't think Vars are thread-local. They're one of the shared
mutable state primitives. They can be defacto thread local if only
used by a single thread but you need a "sufficiently smart compiler"
to notice that.
Hotspot definitely is smart enough in some cases, but I think for
Escape Analysis you currently need a black magic command line
parameter. I'm playing around with: "-XX:+DoEscapeAnalysis
-XX:+UseBiasedLocking" with inconsistant results.
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google
Groups "Clojure" group.
To post to this group, send email to clojure@googlegroups.com
Note that posts from new members are moderated - please be patient with your
first post.
To unsubscribe from this group, send email to
clojure+unsubscr...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/clojure?hl=en
-~----------~----~----~----~------~----~------~--~---
