Investigate ways to avoid forcing fair policy on instrumented ReentrantLocks
----------------------------------------------------------------------------
Key: CDV-742
URL: https://jira.terracotta.org/jira//browse/CDV-742
Project: Community Development
Issue Type: New Feature
Components: Byte Code Transform
Affects Versions: 2.6-stable1
Reporter: Alex Miller
Assignee: Issue Review Board
Currently, we modify ReentrantLock to force the use of fair lock policy as our
instrumented version extends from the fair lock object. This sucks as the
performance of instrumented unclustered locks will be significantly impacted by
this change (I've seen 100x difference in microbenchmarks I wrote for other
purposes).
In code review today we kicked around some ideas for fixing this. Since we
don't allow a lock to become clustered while it's locked, we should be
guaranteed that no one is locking the lock and may be able to make this switch
at the time it becomes clustered instead of at construction time. Another idea
was to actually not extend from the FairSync for our own support since we kind
of supersede the behavior in our own version anyway.
--
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
