RFR(S): 8236035: refactor ObjectMonitor::set_owner() and _owner field setting

David Holmes david.holmes at oracle.com
Tue Jan 28 23:18:18 UTC 2020

On 29/01/2020 7:37 am, Kim Barrett wrote:
>> On Jan 28, 2020, at 12:11 PM, Daniel D. Daugherty <daniel.daugherty at oracle.com> wrote:
>> In summary, I'm planning to leave these two pieces of code as they are
>> in this version of the fix.
>> Kim, please confirm that you are okay with this or not.
> I really dislike mixing acquire/release with {load,store}{load,store}
> for various reason.  I think we should consider eliminating loadload
> and friends in favor of consistent acquire/release usage.  But that's
> a discussion that is clearly out of scope for this change, so I'll
> accept the change as is.

"Horses for courses" as the saying goes. :) acquire/release is a 
higher-level semantics compared to basic loadstore and friends. 
Sometimes all you need are simple ordering constraints, not full 
acquire/release. And sometimes you need a fence.


More information about the hotspot-runtime-dev mailing list