RFR: 8071462: Remove G1ParGCAllocator::alloc_buffer_waste
Stefan Johansson
stefan.johansson at oracle.com
Tue Apr 21 08:49:02 UTC 2015
Thanks Michail,
The change looks good.
Please note that this fix needs to wait on
https://bugs.openjdk.java.net/browse/JDK-8078193 to be sorted out,
otherwise we'll get problems when doing that backout.
Thanks,
Stefan
On 2015-04-20 17:15, Michail Chernov wrote:
> Stefan,
>
> CC'ed you, because message was sent to wrong address, sorry!
>
> Thanks,
> Michail
>
> On 20.04.2015 18:11, Michail Chernov wrote:
>> Hi Thomas, Stefan,
>>
>> Thanks for reviewing!
>>
>> New change is:
>> http://cr.openjdk.java.net/~eistepan/~mchernov/8071462/webrev.04/
>> and its incremental change:
>> http://cr.openjdk.java.net/~eistepan/~mchernov/8071462/webrev.04.inc/
>>
>> Also performed change according to Stefan's comment, extra comma was
>> removed.
>>
>> Thanks,
>> Michail
>>
>> On 20.04.2015 14:06, Thomas Schatzl wrote:
>>> Hi Michail,
>>>
>>> On Fri, 2015-04-17 at 20:46 +0300, Michail Chernov wrote:
>>>> Sorry, typos have been fixed in the message:
>>>>
>>>> Webrev:
>>>> http://cr.openjdk.java.net/~dpochepk/~mchernov/8071462/webrev.03/
>>>>
>>>> Methods were renamed, removed extra comma, class G1PLABWasteStat was
>>>> removed, now G1PLAB::waste(size_t& wasted, size_t& undo_wasted) gets 2
>>>> references. Also I merged my changes with new changes in repo because
>>>> ParGCAllocBuffer and G1ParGCAllocBuffer were renamed to PLAB and
>>>> G1PLAB
>>>> and patch does not apply clear during merging.
>>> looks good, except that the latest change does not compile without
>>> precompiled headers. The problem is the use of
>>> CollectedHeap::fill_with_object() in plab.hpp.
>>>
>>> Please move the methods add_undo_waste(), undo_last_allocation() and
>>> undo_allocation() into the cpp file (only add_undo_waste() is required
>>> to, but that code is not performance critical at all and it is
>>> better to
>>> not have so much stuff in the hpp file) and add the required include
>>> gc_interface/collectedHeap.hpp.
>>>
>>> (Yes, the include is already missing and probably transitively included
>>> through something else, but it is always good to have the includes
>>> complete).
>>>
>>> Thanks,
>>> Thomas
>>>
>>>
>>>
>>>
>>
>
More information about the hotspot-gc-dev
mailing list