Re: [boost] Re: No more tree_traits<> dependencies under placed_ptr<>

From: Larry Evans (jcampbell3_at_[hidden])
Date: 2002-10-03 13:39:43

Philippe A. Bouchard wrote:
 > "David B. Held" <dheld_at_[hidden]> wrote:
 >
 >>"Philippe A. Bouchard" <philippeb_at_[hidden]> wrote:
 >>
[...]

 > I'm ready in one hand, but on the other garbage collection is not
 > yet.  I'm still persuated consecutive deallocations could be

What are "consecutive deallocations"?  Do you mean to accumulate the
deallocations tasks to a certain threshold and then execute them all
at once?

 > optimized also but I do not know the best way to implement it.

If you mean the threshold idea, then maybe lazy local mark scan would
be what you want.  This is like the "eager local mark scan" in the
code I announced in my 9/21 post except that it "accumulates" the
smart pointers in a queue up to a certain threshold and then collects
those that are actually garbage.

I coded a version of this using an older internal pointer enumeration
method and posted it under the now deleted
GarbageCollectionByTemplateSpecialization directory.  I could
resurrect it for you, but, as I said, it's based on and older version
of internal pointer enumeration and is undoubtedly slower, at least
for the enumeration (because it pushes the smartpointers on a stack
instead of using specialized iterators).

 >
 > What about the others?
 >

Based on the lack of response to my other postings about collecting
cycles, I have my doubts whether boost would be interested in yet
another smart pointer.  I could be wrong.