sohi@rsch.wisc.edu (Guri Sohi) (08/12/86)
Does anybody have any information (or pointers to information) about the data stack on the Symbolics 3600? Apparantly it is supposed to reduce garbage collection overhead. Does anybody knows how it works? Guri Sohi sohi@amul.wisc.edu
stern@princeton.UUCP (08/13/86)
[]
You should check out the "Symbolics Technical Summary", which is
available from your local Symbolics sales office. It contains very
well-written descriptions of the various hardware components of the
3600 family machines.
In addition, David Moon (who designed the garbage collection hardware)
has two interesting papers that go into much more detail:
David A. Moon, "Garbage Collection in a Large Lisp System",
Proceedings 1984 ACM Symposium on Lisp and Functional
Programming, pp 235-246
David A. Moon, "Architecture of the Symbolics 3600", Proceedings
12th Annual Symposium on Computer Architecture, pp. 76-83
(1985).
--Hal Stern
Princeton University
{ihnp4, seismo, allegra}!princeton!sternllames@uicsg.UUCP (08/16/86)
>> Does anybody have any information (or pointers to information) >> about the data stack on the Symbolics 3600? ... >> Guri Sohi, sohi@amul.wisc.edu > You should check out the "Symbolics Technical Summary" ... > In addition, David Moon has two interesting papers ... > Hal Stern, {ihnp4, seismo, allegra}!princeton!stern The use of the data stack is described in Volume 8 (Internals, Processes, and Storage Management) of the Symbolics documentation set, pp. 26-28. In summary, special forms are provided that allow you to cons arrays on the data stack, rather that in garbage-collected heap storage. The arrays have dynamic extent (and either lexical or indefinite scope). They get popped off when the form is exited. This facility is like that which lets you cons lists on the control stack. Currently, only arrays can be allocated on the data stack. I don't know if Release 7 will permit more kinds of objects. Rene Llames llames%uicsg@a.cs.uiuc.edu ...!{pur-ee,ihnp4,convex}!uiucdcs!uicsg!llames