RFR JDK-8060017 G1: Report heap sizing time

[Eric Caspole]

Thanks Mikael!

I still need one more review, anyone?
Thanks,
Eric


On 10/13/2015 06:02 AM, Mikael Gerdin wrote:
> Hi Eric,
>
> On 2015-10-12 15:35, Eric Caspole wrote:
>> Hi Mikael,
>> I've returned the expand time as an out parameter in this new webrev:
>>
>>   http://cr.openjdk.java.net/~ecaspole/JDK-8060017/06/webrev/
>>
>> Let me know what you think.
>
> I think this is the most appropriate solution at this point at least.
> I know people are not particularly fond of "output parameters" but I'd
> prefer them to having the expansion code logging to the phase times object.
>
> Reviewed.
> /Mikael
>
>> Thanks,
>> Eric
>>
>>
>>
>> On 09/25/2015 04:33 AM, Mikael Gerdin wrote:
>>> Hi Eric,
>>>
>>> On 2015-09-24 22:44, Eric Caspole wrote:
>>>> Hi everybody,
>>>>
>>>> Could you review this small change to add heap expand time from the VM
>>>> thread into the PrintGCDetails for G1 -
>>>>
>>>>   https://bugs.openjdk.java.net/browse/JDK-8060017
>>>>
>>>> webrev:
>>>>   http://cr.openjdk.java.net/~ecaspole/JDK-8060017/05/
>>>
>>> I'd prefer it if you made the expand time an "out parameter", so that
>>> any caller can get the time measurement. The caller could then decide to
>>> store the time in the phase times structure or not.
>>>
>>> I've had lots of issues with the G1 code where I want to call a function
>>> but doing so causes side effects such as timing measurements.
>>>
>>> Thanks
>>> /Mikael
>>>
>>>>
>>>> The heap expand time taken on the gc threads that would happen during
>>>> object evacuation is already accounted for into the "ObjCopy" time
>>>> although certainly much other work also happens in that phase. But at
>>>> this point the heap expand time on the VM thread is lost into
>>>> unaccounted "Other" time. I started this a long time ago and forgot
>>>> about it due to other work.
>>>> Thanks,
>>>> Eric
>>>
>>>
>