[gclist] GCs Ever-Implemented on Distributed Memory Machines

Marc Shapiro shapiro@prof.inria.fr
Wed, 31 Jul 1996 14:24:49 +0200


 || From: tau@is.s.u-tokyo.ac.jp
 || Date: Sun, 28 Jul 96 22:56:15 +0900
 || Subject: [gclist] GCs Ever-Implemented on Distributed Memory Machines
 || 
 || Dear GC Experts:
 || 
 || I'm interested in garbage collectors which have been implemented and
 || robustly working on distributed memory machines.

We have implemented garbage collection on top of a distributed shared memory,
which represents the worst case of a distributed-memory machine.  It works.
It is not quite ready to be distributed to the world at large, but we have
solved the main theoretical and practical problems.  Take a look at our Web
page
        http://www-sor.inria.fr/SOR/projects/larchant

If you are interested I can send a paper that is not yet accessible via the
Web.

 || I also appreciate if you could comment on my above conjecture (i.e.,
 || implementing GCs on DM machines are rarely completed, because they're so
 || complex).

They pose performance problems which can be solved only by looking at the
theoretical issues involved, which appear quite complex at first glance.  I
think that we have come up with simple and operational solutions, but it was a
lot of hard work.

                                                Marc