RFR: 8137022: Concurrent refinement thread adjustment and (de-)activation suboptimal [v3]

[Kim Barrett]

On Sat, 24 Sep 2022 13:50:47 GMT, Albert Mingkun Yang <ayang at openjdk.org> wrote:

>> Kim Barrett has updated the pull request incrementally with three additional commits since the last revision:
>> 
>>  - wanted vs needed nomenclature
>>  - remove several spurious "scan"
>>  - delay => wait_time_ms
>
> src/hotspot/share/gc/g1/g1ConcurrentRefineThreadsNeeded.cpp line 66:
> 
>> 64:   double alloc_bytes_rate = alloc_region_rate * HeapRegion::GrainBytes;
>> 65:   if (alloc_bytes_rate == 0.0) {
>> 66:     _predicted_time_until_next_gc_ms = 0.0;
> 
> I don't get why time-till-next-gc is zero when the alloc rate is at its minimal -- by continuity, I'd expect this to be the max of time-till-next-gc, i.e. `one_hour_ms`.

A zero value for the prediction indicates that we don't have a valid
prediction, because we're starting up the VM and don't have data yet. That
puts us in a place where we can't really do much with the predictive model.
Later clauses (no refinement needed, or time is short) also "handle" this
case, though maybe there should be some commentary about that.

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

PR: https://git.openjdk.org/jdk/pull/10256