Path: utzoo!utgpu!jarvis.csri.toronto.edu!rutgers!usc!wuarchive!uunet!auspex!guy
From: guy@auspex.auspex.com (Guy Harris)
Newsgroups: comp.arch
Subject: Re: IBM RISC System/6000  AS/400 (actually a diversion on the topic)
Message-ID: <2989@auspex.auspex.com>
Date: 5 Mar 90 19:47:27 GMT
References: <10901@june.cs.washington.edu> <2984@auspex.auspex.com> <10947@june.cs.washington.edu>
Distribution: usa
Organization: Auspex Systems, Santa Clara
Lines: 26


 >:OK, so is VMC code that, in executable form, is in IMPI?
 >
 >    Yes.  Much of it is essentially OS code, although it's part of the
 >    published architecture.  All VMC runs at the IMPI level.

So what language is the OS written in?  Assembler (i.e., something that
translates fairly one-to-one to IMPI)?  RPG or Cobol or one of the other
languages that compiles into MI, with the MI then being statically
compiled into IMPI?

 >    The problem with breaking it down this way is that IMPI can essentially
 >    be thought of as a moving target.  The MI is the published architecture,
 >    and all future AS/400's will conform to MI.  However, the IMPI can change
 >    quite radically.  All programs that are written at MI level will continue
 >    to run, even if they were translated from MI to IMPI (the os will
 >    take care of automatically retranslating them).

So you then think of System/38 and AS/400 as being a set of machines
with potentially binary-incompatible instruction sets, but that all run
compatible interpreters/MI compilers....  Perhaps this is sort of like
moving Smalltalk bytecodes between different machines with different
underlying processors.

So how come if the IMPI *can* change, it *hasn't* changed?  (*Did* they
write the OS in assembler language? :-))