Path: utzoo!attcan!uunet!lll-winken!lll-tis!helios.ee.lbl.gov!pasteur!ames!ncar!ico!rcd
From: rcd@ico.ISC.COM (Dick Dunn)
Newsgroups: comp.arch
Subject: Re: Self-modifying code
Summary: good, bad, ugly
Keywords: 6600
Message-ID: <7377@ico.ISC.COM>
Date: 19 Jul 88 20:02:02 GMT
References: <5254@june.cs.washington.edu> <76700032@p.cs.uiuc.edu> <361@scolex> <7362@ico.ISC.COM>
Organization: Interactive Systems Corp, Boulder, CO
Lines: 16


One other quick note about self-modifying code on the CDC 6000/7000 series:
It was generally a very bad idea to do it because of the instruction
lookahead.  The 7600 looked farther ahead than the 6600, and the 6400 only
had a one-word instruction lookahead.  Thus, depending on the processor,
it might or might not have already fetched an instruction before you
modified it--leading to the superficially baffling phenomenon that the
processor would appear not to have executed the instructions which would
show up in a memory dump.

But it's an ill wind, etc...  The fact that self-modifying code behaved
differently (but predictably so) on different processor models was used to
create a code sequence which could identify the processor at run time and
let code adapt itself accordingly.
-- 
Dick Dunn      UUCP: {ncar,nbires}!ico!rcd           (303)449-2870
   ...Are you making this up as you go along?