-
Bug
-
Resolution: Won't Fix
-
P4
-
None
For a System.gc() the call to ConcurrentMark::abort() can be excessively long.
Cut-and-pasted from a log (some parts left off). I added timers to get these
numbers.
285.569: #0: [Full GC (System.gc()), 6.3754329 secs]
285.569: #0: [Abort concurrent mark (System.gc()), 5.4316610 secs]
285.569: #0: [_nextMarkBitMap->clearAll() (System.gc()) , 2.6187331 secs]
288.188: #0: [clear_all_count_data() (System.gc()) , 2.8120570 secs]
Cut-and-pasted from a log (some parts left off). I added timers to get these
numbers.
285.569: #0: [Full GC (System.gc()), 6.3754329 secs]
285.569: #0: [Abort concurrent mark (System.gc()), 5.4316610 secs]
285.569: #0: [_nextMarkBitMap->clearAll() (System.gc()) , 2.6187331 secs]
288.188: #0: [clear_all_count_data() (System.gc()) , 2.8120570 secs]
- duplicates
-
JDK-8151069 Parallelize clearing the per-thread concurrent mark data structures
- Closed
- relates to
-
JDK-8151215 Modify layout of (large) Concurrent Mark data structures
- Closed