RFR (S): Full GC mark without liveness stats

[Aleksey Shipilev]

On 11/02/2016 12:09 PM, Roman Kennke wrote:
> you are correct, we don't need liveness counting during full-gc.

Excellent!

> I'm not sure I like all the bools all over the place. But I don't see a
> much better way for it either. :-)

Yeah. If that wasn't a performance-sensitive code, we would just push
the bool field. But having in mind the conc GC runs mark code very hot,
template arguments is our sanest option.

I had cleaned up the handling a bit: added the experimental flag and put
the comments in a few critical places. Ok to go?

http://cr.openjdk.java.net/~shade/shenandoah/fullgc-no-live-stats/webrev.02/

> 10% makes me wonder if we can somehow speed up liveness counting. E.g.
> we don't strictly need the liveness counters to be 100% correct. A good
> estimate should be good enough. Dunno if that could help us?

This can be (and would be) handled separately. No matter how fast we are
at liveness counting, completely avoiding it is unbeatable.

Thanks,
-Aleksey