RFR: JDK-8262068: Improve G1 Full GC by skipping compaction for regions with high survival ratio [v12]

[Hamlin Li]

On Tue, 30 Mar 2021 14:07:15 GMT, Thomas Schatzl <tschatzl at openjdk.org> wrote:

>> Hi Thomas, Thank you so much for helping investigate the issue, I will check it tomorrow too.
>
> [Here](https://github.com/tschatzl/jdk/tree/bot-fixes) is a potential fix for this. This needs some discussion on whether an alternatives would be better or not; more of them could be:
> * just initialize a dummy BOT for not-compacted young region
> * exclude young regions from this dead-wood scheme
> In any case this fix should probably go in separately.

I think the first solution is better. Reasons are related to performance:
1. any way, the bot will be rebuild for this young region ( if this young region is compacted, bot will be rebuild when cross_threshold; )
2. 1st solution avoid the copying of this young region.

> [Here](https://github.com/tschatzl/jdk/tree/bot-fixes) is a potential fix for this. 

Thanks for the fix.
>   In any case this fix should probably go in separately.

Do you mean we push this pr, then discuss the fix in another thread? If yes, I will file a bug to track the issue and initialize discussion on this bug.

-------------

PR: https://git.openjdk.java.net/jdk/pull/2760