Wholey@CMU-CS-C.ARPA (01/08/84)
From: Skef Wholey <Wholey@CMU-CS-C.ARPA> Date: Sun, 18 Dec 83 13:57 EST From: Steven Gutfreund <gutfreund%umass-cs@CSNet-Relay> I do find the Symbolics to be slower than a stand-alone PDP-11/70 running RSTS. I will not state that this is solely do to demand paging, but my inclination runs that way. However one possible alternate explanation is that the FSedit, Dired, and Zmacs code is not poured over and optimized the way this system code is in the DEC operating systems groups. You admit that on a PDP-11 running RSTS, you can't FSEdit, Dired, or run an editor as sophisticated as Zmacs. Something to keep in mind is that one can put a LOT of physical memory into a 3600 (30 megabytes). How much physical memory does your 3600 have? 2 Meg is the minimum configuration, and with only 2 Meg, switching between the applications does indeed cause a lot of paging. You don't have to give up quick response time for great user-interface if you're willing to shell out a few more kilobucks for extra memory. The machines that Symbolics uses for demos are filled to the gills with memory, and look really FAST. Demand paging isn't necessarily the time-eater that many of you are making it out to be. Many computer manufacturers are building machines with not nearly enough physical memory to efficiently run certain applications. Clever programming techniques like the initialize-while-we-read-the-code-from-disk method suggested can be used for a small number of programs, but it just isn't cost effective to do this sort of thing for EVERY program. If you want your Lisp Machine to run fast, buy more memory. That IS an option. If you have a demand-paged machine with a puny physical address space and can't give it more memory, say nasty things about the guy who designed it. Demand paging and super-winning programming environments can be fast if there's enough memory to run in.