RFR: Traversal: Initialize extra liveness arrays when conc-threads > par-threads

[Roman Kennke]

Let's put that here for reference:

On 02/22/2018 01:25 PM, Roman Kennke wrote:
> I just checked why sh-conc-mark does not suffer the same problem.
> There, we init the liveness arrays at top of mark loop, and flush it
> at the end. This way, we can never get a mismatch of workers. We
> should just do the same:
>
> http://cr.openjdk.java.net/~rkennke/init-extra-liveness/webrev.02/

Wrong RFR thread, but patch looks good.

On Thu, Feb 22, 2018 at 12:21 PM, Aleksey Shipilev <shade at redhat.com> wrote:
> On 02/22/2018 12:19 PM, Roman Kennke wrote:
>> http://cr.openjdk.java.net/~rkennke/init-extra-liveness/webrev.01/
>
> I'd extend the existing block that has:
>   ShenandoahGCPhase phase_prepare(ShenandoahPhaseTimings::traversal_gc_prepare);
>
> ...to cover this new path (and probably RP setup too) to monitor if that was a good idea.
>
> Otherwise looks good.
>
> Thanks,
> -Aleksey
>