AtomicLong/AtomicInteger (and types that depend on them) will be slow and
concurrency bottlenecks in IBM JDK
------------------------------------------------------------------------------------------------------------
Key: CDV-694
URL: https://jira.terracotta.org/jira//browse/CDV-694
Project: Community Development
Issue Type: Bug
Components: Byte Code Transform
Affects Versions: 2.5.4, 2.6-stable0
Reporter: Tim Eck
Assignee: Issue Review Board
AtomicLong and AtomicInteger have special class adapters for use in the IBM
JDK. From what I can gather, trying to add the Manageable interface (and/or the
$__tc_managed field) causes seg faults in those classes. To work around the
problem, these types end up using the client object manager to determine if
they are managed -- that check is a bottleneck.
This is an example stack trace coming though a Random (which uses AtomicLong):
"worker-100" prio=1 tid=0x0000002c302fc820 nid=0x345c waiting for monitor entry
[0x0000000049243000..0x0000000049243bb0]
at
com.tc.object.ClientObjectManagerImpl.basicLookup(ClientObjectManagerImpl.java:852)
- waiting to lock <0x0000002adf83ba38> (a
com.tc.object.util.IdentityWeakHashMap)
at
com.tc.object.ClientObjectManagerImpl.lookupExistingOrNull(ClientObjectManagerImpl.java:394)
at
com.tc.object.ClientObjectManagerImpl.isManaged(ClientObjectManagerImpl.java:451)
at com.tc.object.bytecode.ManagerImpl.isManaged(ManagerImpl.java:694)
at com.tc.object.bytecode.ManagerUtil.isManaged(ManagerUtil.java:385)
at com.tcclient.util.DSOUnsafe.compareAndSwapLong(DSOUnsafe.java)
at
java.util.concurrent.atomic.AtomicLong.compareAndSet(AtomicLong.java:110)
at java.util.Random.next(Random.java:141)
at java.util.Random.nextDouble(Random.java:380)
at java.lang.Math.random(Math.java:694)
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
https://jira.terracotta.org/jira//secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
_______________________________________________
tc-dev mailing list
[email protected]
http://lists.terracotta.org/mailman/listinfo/tc-dev
