RFR (S): 8202842: G1 footprint regressions in jdk11+10

Stefan Johansson stefan.johansson at oracle.com
Wed May 23 11:29:37 UTC 2018



On 2018-05-16 17:41, Thomas Schatzl wrote:
> Hi Stefan,
>
> On Wed, 2018-05-16 at 12:11 +0200, Stefan Johansson wrote:
>> Hi Thomas,
>>
>> On 2018-05-16 11:03, Thomas Schatzl wrote:
>>> Hi all,
>>>
>>>     can I have reviews for this change that fixes (and actually
>>> improves upon the baseline) some startup footprint regression
>>> introduced by changes in jdk10+10?
>>>
> [...]
>>> CR:
>>> https://bugs.openjdk.java.net/browse/JDK-8202842
>>> Webrev:
>>> http://cr.openjdk.java.net/~tschatzl/8202842/webrev/
>> Looks good, thanks for doing the analysis and fixing this.
>    thanks for your review; while talking about this change internally we
> thought that it would be useful to allow the old behavior where the
> FromCardCache is eagerly backed by memory.
>
> So if AlwaysPreTouch is enabled, G1 will pretouch the G1FromCardCache
> with the following change.
>
> http://cr.openjdk.java.net/~tschatzl/8202842/webrev.0_to_1 (diff)
> http://cr.openjdk.java.net/~tschatzl/8202842/webrev.1 (full)
Still good!
StefanJ
> Thanks,
>    Thomas
>



More information about the hotspot-gc-dev mailing list