When discussing solutions for JDK-8170520, it was suggested that returning humongous chunks to the chunk manager should follow the same paths as non-humongous chunks. In particular, it should be handled by the ChunkManager itself instead of the SpaceManager desctructor. That way code duplication and pitfalls can be avoided and there is no need anymore to expose ChunkManager internals.
See also discussion: http://mail.openjdk.java.net/pipermail/hotspot-runtime-dev/2016-November/021946.html
See also discussion: http://mail.openjdk.java.net/pipermail/hotspot-runtime-dev/2016-November/021946.html
- relates to
-
JDK-8183228 Fix optimized build after JDK-8170933
-
- Resolved
-
-
JDK-8176808 Split up metaspace.cpp
-
- Resolved
-
-
JDK-8166690 Reduce metaspace waste by dynamically merging and splitting metaspace chunks.
-
- Closed
-
-
JDK-8170520 Make Metaspace ChunkManager counters non-atomic
-
- Resolved
-