RFR: 8205559: Remove IN_CONCURRENT_ROOT Access decorator

Kim Barrett kim.barrett at oracle.com
Tue Jun 26 18:26:30 UTC 2018

> On Jun 26, 2018, at 7:50 AM, Stefan Karlsson <stefan.karlsson at oracle.com> wrote:
> Looks good.
> StefanK


> On 2018-06-24 23:51, Kim Barrett wrote:
>> Please review the removal of the IN_CONCURRENT_ROOT Access decorator.
>> All non-AS_RAW IN_NATIVE accesses will be treated as potentially
>> concurrent, and so will require barriers appropriate to the selected
>> collector.
>> Renumbered the Access decorators, eliminating gaps and mis-orderings
>> that had accrued from previous removals and renamings.
>> Fixed StringDedupTable::lookup usage of the Access API.  It was using
>> IN_CONCURRENT_ROOT incorrectly, and was not using AS_NO_KEEPALIVE
>> where it should have been.  There are other places where this class is
>> not using the Access API but should be, involving the _obj field of
>> StringDedupTableEntry; addressing those will be part of a future CR,
>> since currently only ZGC would fail because of these and ZGC does not
>> yet support string deduplication.
>> CR:
>> https://bugs.openjdk.java.net/browse/JDK-8205559
>> Webrev:
>> http://cr.openjdk.java.net/~kbarrett/8205559/open.00/
>> Testing:
>> Mach5 tier1,2,3, hs-tier4,5
>> Aurora perf testing for G1 and Parallel detected no performance regressions.

More information about the hotspot-dev mailing list