[concurrency-interest] @Contended (JEP-142)
dl at cs.oswego.edu
Wed Nov 28 07:09:11 EST 2012
On 11/28/12 04:53, Aleksey Shipilev wrote:
> On 11/28/2012 01:44 PM, √iktor Ҡlang wrote:
>> Would it help adoption if this would get backported to JDK7 (you will
>> have to have the recent update release nevertheless)? Would that be
>> sensible to backport to JDK6 at all?
>> Even if that was done, what about J9, Dalvik etc?
>> I think the sad fact here is that it will essentially only be usable
>> within the JDK for the foreseeable future.
> (I'll let Doug to chime in on J9, Dalvik, etc. endeavors with
> @Contended, if there are any).
I regularly keep in touch with these folks about JVM-support
issues that impact j.u.c. For @Contended though, there's
not much immediate impact: since it is an annotation, it can be
ignored by JVMs until underlying support is in place. And supporting
it is self-motivating. Ignoring @Contended for, say, the
to-be-updated version of LongAdder that will use it rather
than the current manually guestimated padding leads to about
a factor of 40 slowdown on some benchmarks/machines.
More information about the Concurrency-interest