Thanks John.
I was curious about the details so I took a dive in to the source to
see for myself.
In case anyone else stumbles upon this here's what I found:
In Agent.java, the number of worker threads for (send) are defined
like this:
final public static ExecutorService pooledExecutor =
Executors.newFixedThreadPool(2 + Runtime.getRuntime
().availableProcessors());
The clojure (send) calls Java Agent dispatch(), which winds up using
the pooledExecutor.
Clojure (send-off) follows the same path but winds up using the
soloExecutor - which can spawn (and temporarily cache) an unlimited
number of threads as required:
http://java.sun.com/javase/6/docs/api/java/util/concurrent/Executors.html#newCachedThreadPool()
Random thought: Let's test (send) vs (send-off). New results using
(send):
(the -server jvm produces some wild results for a bit then I get
something crazy: 7.9ms):
Clojure=> (agent-speed-test)
"Elapsed time: 1164.702387 msecs"
20001
Clojure=> (agent-speed-test)
"Elapsed time: 1416.682837 msecs"
20001
Clojure=> (agent-speed-test)
"Elapsed time: 7.907515 msecs" <------ ** ???? LOL **
This happens infrequently, but it did happen more than once.
Also, I note that only 2 cores are used on my quad core2 duo. Weird.
It's as if
Runtime.getRuntime().availableProcessors() fails (returns zero?)
--~--~---------~--~----~------------~-------~--~----~
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
-~----------~----~----~----~------~----~------~--~---
