Duplicate :
|
|
Relates :
|
|
Relates :
|
Clearing the per-thread concurrent mark data structures is single threaded. This is a bad idea in case of large heaps, significantly lengthening the concurrent mark cycle. Parallelize this phase.
|