RFR: Don't initially evacuate Synchronizer roots

[Zhengyu Gu]

Okay.

-Zhengyu


On 11/23/2016 10:40 AM, Roman Kennke wrote:
> Am Mittwoch, den 23.11.2016, 10:24 -0500 schrieb Zhengyu Gu:
>> You may want to add assertion to ensure that it is always the case,
>> and prevent the breakage in the future.
> I think we've already got asserts in synchronizer.
>
> We simply don't expect Synchronizer roots to be always in to-space
> anymore. All uses of ObjectMonitor's object do the barrier before using
> it.
>
> Roman
>
>> Thanks,
>>
>> -Zhengyu
>>
>>
>>
>> On 11/23/2016 10:09 AM, Roman Kennke wrote:
>>> I checked all the uses of ObjectMonitor from the Synchronizer
>>> monitor
>>> list. In all the places that they are used we do have proper
>>> barriers.
>>> There is no place that expects the oop in the monitor to always be
>>> to-
>>> space. Therefore we don't need to evacuate those ObjectMonitor
>>> roots at
>>> the init-evac pause.
>>>
>>> Tested with jtreg and SPECjvm2008
>>>
>>> http://cr.openjdk.java.net/~rkennke/noevacsyncroots/webrev.00/
>>>
>>> Ok?
>>>
>>> Roman