-
Enhancement
-
Resolution: Fixed
-
P4
-
17, 21, 22, 23
-
b12
In `PSParallelCompact::mark_obj`, `_summary_data.add_obj` uses atomic-operation to record live bytes info, which hinders scalability during Full GC.
One can take some inspiration from `G1RegionMarkStatsCache` to address this issue.
One can take some inspiration from `G1RegionMarkStatsCache` to address this issue.
- duplicates
-
JDK-8323791 Parallel: Using non-atomic for live bytes update during Full GC
- Closed
- relates to
-
JDK-8320165 Parallel: Full GC code is very slow due to quadratic calc_new_address
- Closed