RFR: 8184181: Use oopDesc::cas_set_mark() instead of raw CAS when accessing oop header
Daniel D. Daugherty
daniel.daugherty at oracle.com
Mon Jul 17 14:23:49 UTC 2017
On 7/17/17 12:44 AM, Roman Kennke wrote:
> Am 15.07.2017 um 00:35 schrieb Daniel D. Daugherty:
>> On 7/14/17 5:31 AM, Roman Kennke wrote:
>>> BiasedLocking and ObjectSynchronizer both use raw Atomic::cmpxchg_ptr()
>>> calls to CAS the header word of an oop. It's cleaner to use
>>> oopDesc::cas_set_mark() instead. Besides this being a cleanup, in the
>>> future (Shenandoah) we will also require GC barriers to do the right
>>> thing there.
>>>
>>> Webrev:
>>> http://cr.openjdk.java.net/~rkennke/8184181/webrev.00/
>>> <http://cr.openjdk.java.net/%7Erkennke/8184181/webrev.00/>
>> General comment
>> - Please update copyright years before pushing as needed.
> I keep forgetting that... thanks for reminding.
>> src/share/vm/runtime/biasedLocking.cpp
>> No comments.
>>
>> src/share/vm/runtime/synchronizer.cpp
>> No comments.
>>
>> Thumbs up!
> Differential:
> http://cr.openjdk.java.net/~rkennke/8184181/webrev.01.diff/
> <http://cr.openjdk.java.net/%7Erkennke/8184181/webrev.01.diff/>
>
> Full:
> http://cr.openjdk.java.net/~rkennke/8184181/webrev.01/
> <http://cr.openjdk.java.net/%7Erkennke/8184181/webrev.01/>
>
> Can I have a sponsor?
I should be able to sponsor this for you.
Dan
>
> Thanks & cheers,
> Roman
>
More information about the hotspot-runtime-dev
mailing list