RFR: Expand old on demand [v63]

[William Kemper]

On Thu, 25 May 2023 03:20:31 GMT, Kelvin Nilsen <kdnilsen at openjdk.org> wrote:

>> This PR describes several proposed changes to dynamically adjust the sizes of old-gen and young-gen.  In general, the objective is to keep old-gen as small as possible so that there is an abundance of memory available for the young-gen allocation runway.
>> 
>> This PR now passes all GHA pre-integration tests and other internal CI tests.
>
> Kelvin Nilsen has updated the pull request incrementally with two additional commits since the last revision:
> 
>  - Allow OOM failures to occur more quickly
>    
>    After a change to increase the defaulit number of consecutive
>    degenerated cycles that can occur before a full gc is triggered,
>    the number of allocation retries that must be completed before throwing
>    an OutOfMemory exception increased significantly.  This commit provides
>    a mechanism to more quickly decide to throw OOM.
>  - Revert default value of ShenandoahGuaranteedYoungGCInterval
>    
>    The previous change was somewhat arbitrary and masks a bigger problem
>    with triggering GC following periods of idle activity.  An alternative
>    approach to triggering should be developed in the future.

I think it would be simpler/fewer changes to pull full gc count of the existing count in `ShenandoahCollectorPolicy`.

src/hotspot/share/gc/shenandoah/shenandoahHeap.hpp line 358:

> 356: 
> 357:   // How many full-gc cycles have been completed?
> 358:   volatile size_t _completed_fullgc_cycles;

`ShenandoahCollectorPolicy` already counts full GC cycles (implicit and  explicit). We could add a `get_full_gc_count` there without duplicating any functionality.

src/hotspot/share/gc/shenandoah/shenandoahHeap.hpp line 397:

> 395: 
> 396:   void finish_fullgc() {
> 397:     Atomic::add(&_completed_fullgc_cycles, (size_t) 1);

`Atomic::inc`? Also - this will only be called on a safepoint at the end of full gc? Does it need atomic access?

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

Changes requested by wkemper (Committer).

PR Review: https://git.openjdk.org/shenandoah/pull/248#pullrequestreview-1444165349
PR Review Comment: https://git.openjdk.org/shenandoah/pull/248#discussion_r1205688396
PR Review Comment: https://git.openjdk.org/shenandoah/pull/248#discussion_r1205690065