On 06/06/12 13:09, Rémi Forax wrote:
I haven't read the whole code (I need a blackboard :)
(The mostly-non-blocking retries here lead to a lot of code sprawl. Macros would help :-)
I wonder if it's not better to use unsafe.tryMonitorEnter() instead of a lock bit.
I need a lock with a form of tryLock, so while I do use built-in monitors and/or AQS for actual locking, I need to preface use with lock bit in regular non-tree-bins.
Also the class depends on LongAdder, is it a requirement or an AtomicLongUpdater can be used ?
If this ever makes it as far as contemplating JDK7 ports, a minimal form of LongAdder could be pulled in as a static inner class. -Doug
