Xref: utzoo comp.sys.amiga.tech:15767 comp.sys.amiga.hardware:4424
Path: utzoo!attcan!uunet!cs.utexas.edu!usc!zaphod.mps.ohio-state.edu!wuarchive!psuvax1!rutgers!cbmvax!daveh
From: daveh@cbmvax.commodore.com (Dave Haynie)
Newsgroups: comp.sys.amiga.tech,comp.sys.amiga.hardware
Subject: Re: RISC Amiga and multiprocessing
Message-ID: <15629@cbmvax.commodore.com>
Date: 6 Nov 90 00:56:31 GMT
References: <1990Nov2.070729.19128@engin.umich.edu> <1990Nov2.190352.5801@jato.jpl.nasa.gov>
Reply-To: daveh@cbmvax.commodore.com (Dave Haynie)
Distribution: na
Organization: Commodore, West Chester, PA
Lines: 55

In article <1990Nov2.190352.5801@jato.jpl.nasa.gov> jdickson@jato.Jpl.Nasa.Gov (Jeff Dickson) writes:

>	This is merely speculation. I find it difficult to believe that 
>CBM could one day dump Motorola CPU's and take on say Intel's. 

As I said before, I don't really expect anything called an "Amiga" to run with
a non-Motorola 680x0 family CPU as it's heart, simply because of software
compatibility issues.  That doesn't imply that other CPUs couldn't find a
useful place in an Amiga; in fact, some alternate CPUs are available today
as plug ins.

>What leads me to this speculation are the major differences in Motorola's 
>and Intel's bus philosophy. For one, Motorola goes for memory mapped I/O 
>while Intel goes for non memory mapped I/O.

That's primarily a feature of Intel 80x86 CPUs; I don't think there are any
RISC CPUs or other new interesting processors, from Intel or anyone else,
that work that way now.  It's not a problem anyway, since any hardware in
special "I/O memory" can be remapped as memory mapped hardware very simply
when you're designing the system.  Or, more to the point, just because you
have special I/O instructions doesn't mean you have to use them -- an 80386
can do memory mapped I/O if you put it in a memory mapped system.  This silly
I/O mapping was really only useful back in the days of 8 bit CPUs; the
special I/O instructions on 8080 and Z-80 let you have all 64K of address
spaced devoted to ROM and RAM, while still letting you at some I/O registers.
Most of the 6502 based systems had memory banking of some kind to overlay 
memory and I/O as needed.

>Another, Motorola has a more robust bus handshaking protocol. Motorola CPU's
>say "here it is" and the type of bus cycle lasts until whoever got it says
>"OK. I got it". Intel CPU's say "Go for it" and its up to who ever it is for 
>to assert wait cycles if more time is needed. 

Again, this is the older 80x86 line, and it's quite simple to convert one type
of cycle structure into another.  I don't think we'll have to worry about
glomping 80386s onto Amiga chips, and most likely, any add-on CPU that's not
a 680x0 family memory belongs on the Zorro III bus, which is relatively
CPU family independent.

>I also believe there are big differences in interupts and interupt levels. 

That's an issue with ISA bus more than 80x86.  The ISA bus used edge sensitive
active high interrupts, which made it impossible to share interrupts between
multiple cards.  Silly, since the level sensitive active low interrupts used
on the Amiga bus, which permit sharing of interrupt lines, were quite common
in the systems that preceeded the first XT and AT bus machines.  I gather that
EISA and MCA bus both allow interrupt sharing.

>Jeff S. Dickson                              jdickson@zook.jpl.nasa.gov


-- 
Dave Haynie Commodore-Amiga (Amiga 3000) "The Crew That Never Rests"
   {uunet|pyramid|rutgers}!cbmvax!daveh      PLINK: hazy     BIX: hazy
	Standing on the shoulders of giants leaves me cold	-REM