RFR (S) CR 8014966: Add the proper Javadoc to @Contended
aleksey.shipilev at oracle.com
Thu May 30 08:55:02 UTC 2013
On 05/30/2013 12:49 PM, Aleksey Shipilev wrote:
> On 05/29/2013 10:48 PM, roger riggs wrote:
>> Has it been considered that highly contended field should not be allocated
>> in the object itself anyway but in some pool or structure better suited
>> to its access characteristics? A read-only indirection could provide
>> some ability to allocate contended locks elsewhere where they would
>> not disrupt the allocation mechanism. For example, in a pool
>> of write-only or write-mostly values for the processor/core/thread.
>> Of course, that has an access cost too.
> That's an interesting idea, and we had discussed it a bit internally.
> The major problem is a mess in runtime to segregate the accesses to the
> plain fields and the contended ones. This change is much more intrusive
> than the smart padding within the instance itself; and the benefits are
> unclear. You will still have to reserve the space at the allocation time
> (yes, sophisticated schemes can do that lazily), you still have
> indirection, etc.
...actually the largest trouble: you have to garbage-collect
off-instance fields somehow, which includes tracking the reachability,
which probably entails write barriers for those fields... Remarkably,
this sounds like another use for ephemerons :)
[---- I'd like to draw the line for metaphysical discussions here ----]
More information about the core-libs-dev