Multiple disjoint object memory in Smalltalk

Noel J. Bergman noel at devtech.com
Tue May 29 02:10:47 UTC 2001


> The key issue when dealing with multiple object memories *is*
> garbage collection. It is because the garbage collector needs
> to know what objects are still alive, where it can move objects
> around and when and where to stop.

Yes, but if all that one wanted to do were to support a segmented heap, one
could adopt one of the garbabe collection schemes that have been worked out
in other virtual machines, such as the scheme Jon Hylands discusses as being
in VisualWorks, or John M McIntosh mentioned for VisualAge.

> I've been talking with Dan about some of this stuff recently
> (the idea was to see if we could provide an OM on a per project
> basis) and we've had some pretty cute ideas (though nothing
> real workable for the moment).

Exactly ... this is what I have been trying to say.  You want multiple
Object Memories that are actually segregated; by project, for example.  This
seems to me to be a more challenging (and thus interesting) issue.  So far,
only John Sarkela's reference to Tensegrity sounds like a might deal with
this subject.  The others all seem to deal with the simpler case of a
segmented heap.

> I suggest we'll both watch the other thread and see
> if somebody has an outstandingly interesting idea ;-)

Hey, I agree with you on that, but first we've got to make sure we're all
talking about the same problem, and not just a segmented address space.

	--- Noel





More information about the Squeak-dev mailing list