RFR (M): 8008920: Tracing events for heap statistics

Mikael Gerdin mikael.gerdin at oracle.com
Mon Mar 25 07:01:53 PDT 2013


Erik,

On 2013-03-25 14:47, Erik Helin wrote:
> All,
>
> I have updated the change based on feedback from Bengt and Jesper.
>
> The differences between webrev.01 and webrev.02 are:
> - Updated the comment in VM_GC_HeapInspection.
> - Updated the label for the object count event.
>
> New webrev located at:
> http://cr.openjdk.java.net/~ehelin/8008920/webrev.02/

In concurrentMarkSweepGeneration.cpp:
It would feel more appropriate to have the call to 
report_object_count_after_gc to before we _collectorState to Sweeping.

The rest looks good to me.

/Mikael

>
> Thanks,
> Erik
>
> On 03/25/2013 12:33 PM, Erik Helin wrote:
>> Bengt,
>>
>> On 03/25/2013 12:28 PM, Bengt Rutisson wrote:
>>>
>>> Hi Erik,
>>>
>>> This looks good.
>>>
>>> One small nit. In VM_GC_HeapInspection::collect() you preserved the
>>> comment from VM_GC_HeapInspection::doit(). The comment talks about
>>> issuing a warning message, but the code to log that message is left in
>>> doit(). Could we update the comment to be more consistent with the new
>>> code?
>>
>> I will update the comment.
>>
>> Thanks,
>> Erik
>>
>>> Thanks,
>>> Bengt
>>>
>>> On 3/22/13 9:55 AM, Erik Helin wrote:
>>>> All,
>>>>
>>>> based on internal feedback, this change has been updated.
>>>>
>>>> The event has been renamed to vm/gc/detailed/object_count_after_gc.
>>>> Furthermore, the part of the change that updated
>>>> heapInspection.cpp/hpp has been extracted to a separate change (also
>>>> out for review on hotspot-gc-dev at openjdk.java.net).
>>>>
>>>> New webrev located at:
>>>> http://cr.openjdk.java.net/~ehelin/8008920/webrev.01/
>>>>
>>>> Thanks,
>>>> Erik
>>>>
>>>> On 03/16/2013 11:38 AM, Erik Helin wrote:
>>>>> Hi all,
>>>>>
>>>>> this change adds an event, vm/gc/detailed/class_count_after_gc, that
>>>>> sends information about the number of instances of each class on the
>>>>> heap (as well as size and the name of the class).
>>>>>
>>>>> The event is sent after the marking phase for all old collections
>>>>> and an
>>>>> object will be counted as live if the GC think that the object is live
>>>>> (this is only tricky in the case of a concurrent collection).
>>>>>
>>>>> Webrev:
>>>>> http://cr.openjdk.java.net/~ehelin/8008920/webrev.00/
>>>>>
>>>>> Bug:
>>>>> http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=8008920
>>>>>
>>>>> Testing:
>>>>> JPRT
>>>>>
>>>>> Thanks,
>>>>> Erik
>>>>
>>>
>>
>


More information about the hotspot-gc-dev mailing list