I can neither confirm nor deny that I am working on an incremental
multiprocessor garbage collector without any stop-the-world requirements.
It is true that in the implementation that might or might not exist, processor
cache synchronization is performed between the point where the object becomes
unreachable and the point where the finalizer is run. However, the GC
algorithm is designed to greatly reduce synchronization requirements. Future
implementations could use instructions (which might or might not exist) which
only flushed certain cache lines, but such an optimization would be rendered
impossible by Bill's proposed finalizer semantics.
Of course, I can neither confirm nor deny that future IBM microprocessors will
have such instructions.
JavaMemoryModel mailing list - http://www.cs.umd.edu/~pugh/java/memoryModel
This archive was generated by hypermail 2b29 : Thu Oct 13 2005 - 07:00:24 EDT