RFR: 8272083: G1: Record iterated range for BOT performance during card scan [v4]

[Stefan Johansson]

On Sat, 9 Oct 2021 10:05:07 GMT, Yude Lin <duke at openjdk.java.net> wrote:

>> A fix to the problem in 8272083 is to use a per-worker pointer to indicate where the worker has scanned up to, similar to the _scanned_to variable. The difference is this pointer (I call it _iterated_to) records the end of the object and _scanned_to records the end of the scan. Since we always scan with increasing addresses, the end of the latest object scanned is also the address where BOT has fixed up to. So we avoid having to fix below this address when calling block_start(). This implementation approximately reduce the number of calls to set_offset_array() during scan_heap_roots() 2-10 times (in my casual test with -XX:G1ConcRefinementGreenZone=1000000).
>> 
>> What this approach not solving is random access to BOT. So far I haven't found anything having this pattern.
>
> Yude Lin has updated the pull request incrementally with four additional commits since the last revision:
> 
>  - Fixed a bug in array claim
>  - Fixed a bug that causes premature deactivate
>  - Renaming variables
>  - Renaming

A bit late to the party. Here are some thoughts from me. 

First of all, many thanks for identifying this issue. Certainly something we need to address. 👍 

Secondly, did you ever explore fixing the bot as we go in the evacuation phase? And if so, have you seen any issues with such an approach? I did a quick PoC (that doesn't yet handle out of PLAB allocations) and from my initial measurements it looks quite promising:
https://github.com/openjdk/jdk/compare/master...kstefanj:rnd-bot-yc-update

Pushing this work out of the pause might be the end goal, but analyzing the overall affect of increased concurrent work is harder than just shuffling around the work in the pause. My initial measurements with the above shows shorter scan-times at the cost of slightly longer copy-times, but the overall pause is shorter.

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

PR: https://git.openjdk.java.net/jdk/pull/5039