RFR: Refactor region flags into FSM

[Aleksey Shipilev]

On 08/24/2017 03:09 PM, Roman Kennke wrote:
> + if (region->is_allocatable() && !heap->in_collection_set(region)) { > If it's allocatable, it should not be in the cset?

Yes. Fixed. There is one other place where checking against cset collection is sensible, because
region flags are not yet set. But here, we can rely on flags.

Also renamed is_allocatable() -> is_alloc_allowed().

> Also, the heap cset table should eventually also go elsewhere (HeapRegionManager or however you
> want to call it?)
Yes. Whatever it is called, it should sync up the region flags and cset bitmap.
ShenandoahCollectionSet does it now (crudely), and it should get better.

> Can we have a better name for 'dirty' ? I did that
> way back when, and I was lazy, but now kinda regret it. It doesn't
> really say what it is. 

Yes, good call! heap_region_iterate thinks "dirty" is "cset". Renamed to "Trash", as discussed.

> + if (r->is_empty()) { // TODO: Should be is_allocatable here, to capture regular regions too? 
> I think so :-) 

This is the left-over from Christine's LRU/Generational work. We have discussed it briefly, and
agreed that we can add regular regions there too. But, I want to have it in a separate changeset
anyway, in case of regressions.

New webrev:
  http://cr.openjdk.java.net/~shade/shenandoah/region-fsm/webrev.02/

Still passes hotspot_gc_shenandoah.

Thanks,
-Aleksey