On Thu, Dec 21, 2006 at 03:18:09PM +0100, Mark Wielaard wrote: > Hi Christian, > > On Thu, 2006-12-21 at 15:07 +0100, Christian Thalinger wrote: > > An eclipse-3.2.1 startup-and-shutdown: > > > > 3808 class loads > > 10 classes have a finalizer > > 4 of them are empty > > > > Make your own decision :-) > > Thanks for those stats! But aren't we actually interested in the number > of instances of those 4 classes with empty finalizers (as percentage of > object instances with and without finalizers)? I assume that the > accounting and processing is per instance of such classes that are > garbage collected. Then again I clearly have never written a garbage > collector, especially not one that has to take into account all the > tricks of java finalization. You're absolutely right! This is actually interesting: 3723 of 1171260 (w/: 4300, w/o: 1166960) (these are heap allocations: empty finalizer, overall, with, and without a finalizer) 87% precent of the heap allocations with a finalizer have an empty finalizer. The class responsible for this is java/util/zip/Inflater. > > Is there any measurable runtime difference between the version that does > and doesn't do empty-finalizer-erasure? Have to do some benchmarks first... - twisti
