[concurrency-interest] The Atomic*FieldUpdater situation

Doug Lea dl at cs.oswego.edu
Sun Jul 15 06:37:38 EDT 2012


On 07/14/12 14:01, David M. Lloyd wrote:
> What is the purpose of the access-time access check in the atomic field updater
> classes?
> ...
>
> Is there any recourse to solve this issue?

So to summarize for the moment: Whenever this issue comes up,
there seem to be some upcoming options that will someday
provide good solutions. Along with some shorter-term bandaids that
would make them a little better in some cases. We always end up
not applying the bandaids because nearly all the people impacted
by the performance issues are developing infrastructure-level
components, and know how to correctly encapsulate "Unsafe" bypasses.
This is ugly and uncomfortable, but still probably better
than alternatives. I think the main downside is that people
not in that category reading source code can get the impression
that using Unsafe directly for atomic and fenced accesses is a
recommended technique rather than an act of desperation.

-Doug





More information about the Concurrency-interest mailing list