RFR(M): 6921087: G1: remove per-GC-thread expansion tables from the fine-grain remembered sets

[Bengt Rutisson]

Hi John and Thomas,

This looks good to me.

One minor thing:

Not strictly related to this change, but why do we have 
SAMPLE_FOR_EVICTION defined as 1? I guess this means that it is always 
been on and thus we won't get any performance regressions from removing 
the #if SAMPLE_FOR_EVICTION guards. So, why don't we just go ahead and 
do that? It would simplify the code a bit.

Maybe this should be done as a separate change. It just looks kind of 
strange to me.

Bengt

P.S. Copyright year in heapRegionRemSet.hpp... ;-)



On 2012-06-19 21:21, John Cuthbertson wrote:
> Hi Everyone,
>
> Can I have a review for the cleanup changes, supplied by Thomas 
> Schatzl (Johannes Kepler University at Linz), for this CR? The webrev 
> can be found at: http://cr.openjdk.java.net/~johnc/6921087/webrev.0/
>
> Summary:
> Thomas has removed the per thread expansion tables (PosParPRT) and 
> associated, unused, expansion and compaction code along with some 
> additional unused code. This change significantly reduces the code 
> complexity of the fine grained RSet entries. These changes are a 
> prerequisite for some other changes that Thomas has been working on to 
> optimize the freeing of RSets' fine grained entries.
>
> I've looked at the changes and they have a thumbs up from me.
>
> Testing: OpenDS (performed by Thomas), the GC test suite with heap 
> verification enabled (me), and jprt (me).
>
> Thanks,
>
> JohnC