RFR (L): 8219100: Improve do_collection_pause_at_safepoint

[Kim Barrett]

> On Mar 4, 2019, at 6:16 AM, Thomas Schatzl <thomas.schatzl at oracle.com> wrote:
> 
> CR:
> https://bugs.openjdk.java.net/browse/JDK-8219100
> Webrev:
> http://cr.openjdk.java.net/~tschatzl/8219100/webrev
> Testing:
> hs-tier1-8
> 
> Thanks,
>  Thomas

Looks good.

The only "interesting" thing I noticed is that the call to
DerivedPointerTable::clear seems to have been moved significantly,
including now being within the recorded pause time. But I think the
new location makes more sense.

There were also some comments mentioning JDK-7178365 that have been
deleted. I see though that it has been targeted to 13, so it's on our
radar. It only talks about stuff after recording the pause end though,
and it seems like there is similar stuff going on before the captured
pause start time. FWIW, I think excluding verification time from pause
statistics makes some sense, but other parts do seem questionable.