RFR (M): 8034842: Parallelize the Free CSet phase in G1

[Thomas Schatzl]

Hi Derek,

  I posted new webrevs at

http://cr.openjdk.java.net/~tschatzl/8034842/webrev.3/ (full)
http://cr.openjdk.java.net/~tschatzl/8034842/webrev.2_to_3 (diff)

incorporating most of your suggestions. See below for one comment:

On Tue, 2016-05-10 at 19:00 -0400, Derek White wrote:
> Hi Thomas,
> 
> This is an initial pass of mostly minor comments. I'm still getting a
> handle on the bigger picture.
> 
> ----------------------------------
> src/share/vm/gc/g1/heapRegionSet.hpp
>   - Copyrights.
> 
> ----------------------------------
> src/share/vm/gc/g1/g1CollectedHeap.cpp:
> 
> Lines 4733, 4735:
> - Add comment about being array?
> 
> Line 4744:
> - Is "_pre_used" the same as "bytes (or words?) successfully
> evacuated"?
> 

I renamed the member variable.

> Line 4752:
>   - Consider making do_serial_work directly update G1GCPhaseTimes,
> like 
> do_parallel_work() does. Then cleanup up 
> G1FreeCollectionSetTask::work(). Otherwise fix up spacing in 
> G1FreeCollectionSetTask::work()
> 

Due to more recent experience with and where to put time reporting
code, I now kind of dislike doing the timing within the method doing
the actual work since it disallows you to reuse the methods. That bit
me once too often.

I also changed the do_parallel_work() etc. methods to not return the
time taken any more.

Thanks,
  Thomas