Re: GC on multiple cores

[<don...@is...>]

Sam Steingold writes:
 > why didn't you just start a new thread?
Clearly should have, but this was, in part, an answer to the original.

 > > Is there any prospect of "on line" GC - where some processors
 > > can continue to run lisp code while others are doing GC ?
 > probably not.
 > gc moves stuff around, so you cannot access the lisp heap during gc.
I realize this is a problem, but I don't see (yet) how much of one.
I recall reading about on line GC (a long time ago, around the same
time as circularity testing with small storage), and my impression
(similarly foggy) is that moving objects was not necessarily a
problem.  You might remember lisp machine forwarding pointers?  
Are these used in modern lisp (or other) implementations?  It seems to
me that something similar might be implemented with the same sort of
memory management tricks used now to implement write barriers etc.