>Ok. I guess VMs have improved and now the "conventional wisdom" has
>changed. It used to be that synchronization by itself was a performance
>hit, regardless of how it effected other threads. It had something to do
>with how it was implemented in the VM. But this information is at least a
>couple of years old.
Doing this:
for ( ... ) {
synchronized {
blah
}
}
is considerably slower than this:
synchronized {
for ( ... ) {
blah
}
}
Similarly, if you're going to be calling a synchronized method inside an
inner loop, you will save a lot of time by acquiring the lock outside the
loop. Then the synchronization overhead is shortcircuited because the JVM
sees that we already have the lock and just moves on.
--
Brian Goetz
Quiotix Corporation
[EMAIL PROTECTED] Tel: 650-843-1300 Fax: 650-324-8032
http://www.quiotix.com
