RFR: Adaptive thresholds for Traversal

[Roman Kennke]

Am 25.06.2018 um 12:11 schrieb Aleksey Shipilev:
> On 06/25/2018 11:53 AM, Roman Kennke wrote:
>> This 'adopts' (i.e. mostly copy+pastes) the adaptive thresholds stuff
>> from adaptive heuristics to be used in Traversal GC. Not sure if it
>> should somehow be factored to be shared.
>>
>> http://cr.openjdk.java.net/~rkennke/traversal-adaptive-thresholds/webrev.00/
> 
> Looks OK for now.
> 
> This assert is new:
>   167   assert(!heap->has_forwarded_objects(), "no forwarded objects here");
> 
> ...we used to return NONE when forwarded objects are in heap. I assume that never happens outside of
> traversal cycle.
> 

Yeah I was wondering why we care for this situation that should never
actually happen. It's probably a leftover from back when we mixed
traversal/partial with conc-mark. The assert hasn't triggered in
testing, so I guess it's ok.

Roman