Re: A non-sucking garbage collector

FYI: By using a copying garbage collector, which Lieberman recommends for
prototype based systems in his paper anyway, I can get the overhead for my
scheme down to zero, both in  terms of memory and time.

A copying garbage collector does a pass much like a mark  pass but copies
the objects it finds immediately into a new memory area packing them tightly
as it goes.  When it is finished it just deallocates the old memory area
totally.  My new scheme just does the copying from the write-lock routine
also.  Obviously any object being written to is reachable, so it would have
been copied when the GC got to it anyway.

So my new scheme is now zero-overhead and uses the Lieberman-recommended
type of GC.  It is also much simpler to implement than most I have been
looking at.

I'm a very happy puppy.