Re: [boost] An experimental garbage collector for C++ using smart pointers

From: Greg Colvin (gcolvin_at_[hidden])
Date: 2000-10-20 20:01:13

You might also get some good feedback from the gclist mailing list:
   http://iecc.com/gclist/GC-faq.html

I've gotten as far as your readme, and the basic idea appears to be
sound, and very similar to a smart-pointer based collector I published
in the C User Journals a few years ago.  Both have the disadvantage of
requiring a special heap for allocating collected objects.  Two
collectors without that disadvantage, but with troubles of their own,
are to be found on Boost at
   http://www.egroups.com/files/boost/mark_sweep_ptr.zip
   http://www.boost.org/libs/smart_ptr/smarttest.zip
   http://www.boost.org/libs/smart_ptr/smarttests.htm

From: Gary Powell <Gary.Powell_at_[hidden]>
> Hi John,
>   Have you seen http://reality.sgi.com/boehm_mti/gc.html?

This is an excellent collector, widely used and widely ported.  The
current home page is
   http://www.hpl.hp.com/personal/Hans_Boehm/gc/

A commercially supported derivative of this collector is available at
   http://www.geodesic.com.

Either is likely to give performance far better than any pure-C++
solution, at the cost of some system-specific programming.

> There are a bunch of links to other garbage collectors at this site as well.
> (I'm not expert on garbage collecting, unless you count the stuff under the
> sink, I only point to others who claim it.)
> 
>   -gary-
> 
> gary.powell_at_[hidden]
> 
> 
> > -----Original Message-----
> > From: jbandela_at_[hidden] [SMTP:jbandela_at_[hidden]]
> > Sent: Friday, October 20, 2000 5:25 PM
> > To: boost_at_[hidden]
> > Subject: [boost] An experimental garbage collector for C++ using
> > smart pointers
> > 
> > I have posted an experimental garbage collector for C++. It is 
> > written entirely in C++, and uses smart pointers. It uses both 
> > reference counting and and mark-sweep. This is its first submission 
> > and I am sure there are rough edges. I am interested in feedback on 
> > whether this is a viable idea (even if the implementation needs 
> > work), or if it is hopelessly and fundamentally flawed.
> > 
> > Thanks for your time,
> > 
> > John R. Bandela
> > 
> > 
> > 
> > 
> 
> 
> 
>