RFR (S): 8022872: G1: Use correct GC cause for young GC triggered by humongous allocations

Bengt Rutisson bengt.rutisson at oracle.com
Wed Aug 21 13:33:51 PDT 2013


Thanks for the reviews Tony and Thomas!

Bengt


On 8/21/13 9:49 PM, Thomas Schatzl wrote:
> Hi,
>
> On Wed, 2013-08-21 at 21:21 +0200, Bengt Rutisson wrote:
>> Hi Tony,
>>
>> On 8/21/13 6:30 PM, Tony Printezis wrote:
>>> Hi Bengt,
>>>
>>> (surprise!)
>> A nice surprise! But a surprise indeed :)
>>
>>> This is a useful change. Both webrevs look good. I have a slight
>>> preference to the first one. I didn't like the use of
>>> _allocation_failure on the second one here:
>>>
>>>   984       result = do_collection_pause(word_size, gc_count_before,
>>> &succeeded,
>>>   985           GCCause::_allocation_failure);
>>>
>>> given that the allocation didn't really fail - the eden is just full.
>>> The other use of _allocation_failure (in the
>>> VM_G1CollectorForAllocation closure) really implies that an allocation
>>> did fail and G1 is trying to take evasive action.
>> Thanks for looking at this! This was exactly the kind of feedback I was
>> looking for. As I already mentioned I'm also fine with the first webrev
>> and with this explanation I'm also in favor of that one. Unless anyone
>> else has opinions I'll go with that one. Just need one more review...
> Here it is. I agree to keep the _g1_inc_collection_pause gc cause.
>
> Change looks good.
>
> Hth,
> Thomas
>
>



More information about the hotspot-gc-dev mailing list