Unix-Math 101 (Re: Come on wizards)

[Blair P. Houghton]

In article <1991Jan1.144134.18631 at agate.berkeley.edu> shipley at remarque.berkeley.edu (Pete Shipley) writes:
>In article <1990Dec30.004638.12910 at oswego.Oswego.EDU> faex0000 at penelope.Oswego.EDU (Brian (BJT) Talamo) writes:
>Yes,  WHAT WIZARDS!!!!!
>
>What do you want us to write about?

How about "priority calculations I have seen."  The only
two for which I've seen the formulae have been those in
Bach's and Leffler's books, and they don't describe
accurately the priority schemes on any UNIX-alike I've
used, including Ultrix, Umax, SunOS, Aegis/Domain-OS, and
whatever it's called on SGI's Iris workstations.

Most importantly, the "nice-value" (p_nice, set using
nice(1) or renice(8)) has so very little weight in the
actual priority after a very few seconds of CPU time in
vanilla SysV or BSD (see Bach or Leffler to see what I
mean) that it's almost useless as a prioritizing measure.
Are other Unix variants' priority calculations more heeding
of the "nice-value"?

				--Blair
				  "What was it worth and
				   when was it worth that?"