🚀 go-pugleaf

In article <18888@cup.portal.com> ts@cup.portal.com (Tim W Smith) writes:
>If you've got kernel sources for Unix, try hacking it to not
>use pre-emptive multitasking.  It might be interesting to see
>what happens.
>
>The first compute bound process will, of course, hang the system,
>but I bet a lot of things would work well.  Most processes are
>going to give up the CPU quite a lot because they will be doing
>IO, which will lead to a task switch every time they have to
>sleep waiting for disk IO.

We're getting a bit far afield, but this reminds me of the first
computer I was responsible for: a PDP 11/34 running Digital Standard
MUMPS (DSM-11).  DSM-11 was a single-language (MUMPS) operating system
built on top of RSX-11.  MUMPS is a concise interpreter-oriented
language and environment convenient for building multiuser hierarchical
database applications that was (is?) popular in the medical computing
community. 

Anyway... this appeared to be non-preemptive multitasking since a single
compute-bound program did indeed lock the system up solid. (Of course
you can do the same thing to a VAX/VMS system if any CPU hog has a high
enough CPU priority, but MUMPS on a VAX couldn't normally do that.)

My point? Other commercial operating systems from big name vendors have
been successful (somewhat) even with non-preemptive multitasking.
Further, having preemption does not guarantee no CPU starvation.

-Jon

-- 
-----
Jon L Spear: jspear@gryphon.COM    <routing site>!gryphon!jspear
             gryphon!jspear@elroy.jpl.nasa.gov
"With computers we can make billions of mistakes every second!"