[gclist] Baker's treadmill without a doubly linked list.
Henry G. Baker
Wed, 29 Jan 1997 20:44:22 -0800 (PST)
> At 3:24 PM -0800 1/29/97, Henry G. Baker wrote:
> >Isn't this essentially the _original_ (1976-78) RTGC algorithm for
> >'displaced' (CL/Maclisp terminology) objects (i.e., objects with
> >headers) ??
> I don't know. What's the reference? If it is then why was the linked
> list introduced?
Try ftp://ftp.netcom.com/pub/hb/hbaker/RealTimeGC.html (also .ps.Z).
ftp://ftp.netcom.com/pub/hb/hbaker/NoMotionGC.html (also .ps.Z)
This article gives a tiny bit of rationale for the linked list version.
Basically, if you aren't working in machine language, but 'higher level'
languages (Ada, Pascal, C/C++,...); most of these languages make it
pretty difficult to keep things consistent when objects change their
physical location (address). The lists allow the effect of movement,
without actually moving things.