RFR: 8065331: Add JVM trace events for failed allocations

Marcus Larsson marcus.larsson at oracle.com
Thu Apr 9 07:01:52 UTC 2015


On 04/08/2015 06:09 PM, Erik Helin wrote:
> On 2015-04-08, Marcus Larsson wrote:
>> On 04/08/2015 05:13 PM, Erik Helin wrote:
>>> On 2015-04-08, Marcus Larsson wrote:
>>>> Hi again,
>>>>
>>>> During offline discussions with Erik Helin we decided to skip the gc-attempt
>>>> counter altogether, since this information had little value. Also, as the
>>>> "Collect for allocation" event would always be accompanied by the (already
>>>> existing) garbage collection events, this event is unnecessary and has been
>>>> removed as well. Instead, to see allocations pending a particular GC, one
>>>> can look at "Allocations requiring GC" events preceding the garbage
>>>> collection event sent for that GC.
>>>>
>>>> New (significantly smaller) webrev:
>>>> http://cr.openjdk.java.net/~mlarsson/8065331/webrev.02/
>>> Looks good to me, Reviewed.
>>>
>>> Thanks,
>>> Erik
>> Thank you Erik!
> Just had a quick look at the patch file and found this:
>
> @@ -210,7 +210,7 @@
>   class VM_CollectForMetadataAllocation: public VM_GC_Operation {
>    private:
>     MetaWord*                _result;
> -  size_t                   _size;     // size of object to be allocated
> +  size_t                   _size;        // size of object to be allocated
>
> Why did you change the indentation of the comment?

It's a left-over from the larger patch that I seem to have missed. I 
will remove it before pushing. Thanks for catching it!

Marcus

> Thanks,
> Erik
>
>> Marcus
>>
>>>> Thanks,
>>>> Marcus
>>>>
>>>>
>>>> On 04/02/2015 02:55 AM, Marcus Larsson wrote:
>>>>> Hi Jon,
>>>>>
>>>>> Thank you for looking at this.
>>>>>
>>>>> The way I intended the gc count information to be useful is to indicate if/when allocations keep failing. Problematic allocations would then have a gc count higher than 1. It's not really critical that this is the exact total, as it's just an indication, so a partial GC count could work I guess. If you think it would be better I can take a look at how I can reduce the CR by being less precise with this counter.
>>>>>
>>>>> Thanks,
>>>>> Marcus
>>>>>
>>>>> ----- Original Message -----
>>>>> From: jon.masamitsu at oracle.com
>>>>> To: hotspot-gc-dev at openjdk.java.net
>>>>> Sent: Wednesday, 1 April, 2015 4:44:30 PM GMT +01:00 Amsterdam / Berlin / Bern / Rome / Stockholm / Vienna
>>>>> Subject: Re: RFR: 8065331: Add JVM trace events for failed allocations
>>>>>
>>>>> Marcus,
>>>>>
>>>>> These changes are rather extensive and the CR is rather thin
>>>>> on details.  How will the GC count information be useful?
>>>>> Is it critical that it be the total GC count for failed allocation?
>>>>> Could some partial GC count that does not require such
>>>>> extensive changes be almost as useful?
>>>>>
>>>>> Thanks.
>>>>>
>>>>> Jon
>>>>>
>>>>> On 3/30/2015 3:54 PM, Marcus Larsson wrote:
>>>>>> Ping!
>>>>>>
>>>>>> ----- Original Message -----
>>>>>> From: marcus.larsson at oracle.com
>>>>>> To: hotspot-gc-dev at openjdk.java.net
>>>>>> Sent: Tuesday, 24 March, 2015 7:35:08 AM GMT +01:00 Amsterdam / Berlin / Bern / Rome / Stockholm / Vienna
>>>>>> Subject: RFR: 8065331: Add JVM trace events for failed allocations
>>>>>>
>>>>>> Hi,
>>>>>>
>>>>>> Please review the following patch adding new trace events for allocations requiring GC, and GCs triggered by allocation. See the JEP for a description of these two new events.
>>>>>>
>>>>>> CR:
>>>>>> https://bugs.openjdk.java.net/browse/JDK-8065331
>>>>>>
>>>>>> Webrev:
>>>>>> http://cr.openjdk.java.net/~mlarsson/8065331/webrev.00/
>>>>>>
>>>>>> JEP:
>>>>>> https://bugs.openjdk.java.net/browse/JDK-8060092
>>>>>>
>>>>>> Tests:
>>>>>> jprt, jtreg
>>>>>>
>>>>>> Thanks,
>>>>>> Marcus



More information about the hotspot-gc-dev mailing list