RFR (S/M): 8225478: Make G1CMRootRegions independent of HeapRegions
Kim Barrett
kim.barrett at oracle.com
Sat Jun 8 00:31:43 UTC 2019
> On Jun 7, 2019, at 12:52 PM, Thomas Schatzl <thomas.schatzl at oracle.com> wrote:
>
> […]
> CR:
> https://bugs.openjdk.java.net/browse/JDK-8225478
> Webrev:
> http://cr.openjdk.java.net/~tschatzl/8225478/webrev/
> Testing:
> hs-tier1-3 (Sangheon already pushed 8220080 through hs-tier1-8).
>
>
> [0] https://bugs.openjdk.java.net/browse/JDK-8220089
> [1] http://cr.openjdk.java.net/~sangheki/8220089/webrev.4/
> [2]
> http://mail.openjdk.java.net/pipermail/hotspot-gc-dev/2019-May/025774.html
Generally looks good; just a few minor issues and suggestions.
------------------------------------------------------------------------------
src/hotspot/share/gc/g1/g1ConcurrentMark.hpp
271 MemRegion* claim_next();
Perhaps the return type should be const MemRegion* ?
------------------------------------------------------------------------------
src/hotspot/share/gc/g1/g1ConcurrentMark.hpp
315 G1CMRootMemRegions _root_regions;
Member is no longer aligned with all the other members.
(Not that I like that kind of formatting, but it looks odd to have the
one outlier now.)
------------------------------------------------------------------------------
src/hotspot/share/gc/g1/g1ConcurrentMark.hpp
559 // Scan a single root MemRegion which holds from nTAMS to top and mark everything reachable from it.
Maybe drop "which holds from nTAMS to top". The intent is that will
no longer be true because of allocation from survivor regions changing
top. And nTAMS and top are not part of a MemRegion.
------------------------------------------------------------------------------
src/hotspot/share/gc/g1/g1ConcurrentMark.cpp
275 _root_regions = NULL;
Unnecessary.
------------------------------------------------------------------------------
src/hotspot/share/gc/g1/g1ConcurrentMark.cpp
286 assert(start != NULL && end != NULL && start <= end, "Start (" PTR_FORMAT ") should be equal to or greater "
Error message is backward.
------------------------------------------------------------------------------
More information about the hotspot-gc-dev
mailing list