RFR(S) 8154467: Cleanup initialization of GCPolicyCounters

Mikael Gerdin mikael.gerdin at oracle.com
Thu Apr 21 08:16:06 UTC 2016



On 2016-04-20 15:09, Mikael Gerdin wrote:
> Hi Erik,
>
> On 2016-04-20 13:48, Erik Helin wrote:
>> On 2016-04-20, Mikael Gerdin wrote:
>>> Hi all,
>>>
>>> Please review this small cleanup to how GCPolicyCounters are
>>> initialized and
>>> managed.
>>>
>>> GCPolicyCounters* counters() is part of the CollectorPolicy API but
>>> in fact
>>> it is not necessary to expose it at such a general level.
>>> G1 has only used the object internally in the policy code and
>>> managing the
>>> policy counters object can be safely moved inside the G1Policy.
>>>
>>> ParalellGC has ignored initializing the field in the CollectorPolicy at
>>> all, since it does not really use the CollectorPolicy API.
>>>
>>> The Gen collectors can be adapted to access the counters() accessor
>>> through
>>> the GenCollectorPolicy class instead.
>>>
>>> Testing: RBT GC Testing
>>> Webrev: http://cr.openjdk.java.net/~mgerdin/8154467/webrev.0/
>>> Bug: https://bugs.openjdk.java.net/browse/JDK-8154467
>>
>> Nice cleanup. Could you make
>> GenCollectorPolicy::initialize_gc_policy_counters pure virtual and let
>> the implementors of GenCollectorPolicy decide if they want to call
>> ShouldNotReachHere in their implementation?
>
> Sure, I can do that.

I've made the suggested change and agreed with Stefan and Erik that this 
is ready to go in.
Thanks for the reviews!

/Mikael

>
> /Mikael
>
>>
>> Thanks,
>> Erik
>>
>>> /Mikael



More information about the hotspot-gc-dev mailing list