Path: utzoo!attcan!uunet!seismo!sundc!pitstop!sun!quintus!ok
From: ok@quintus.uucp (Richard A. O'Keefe)
Newsgroups: comp.arch
Subject: Re: RISC v. CISC --more misconceptions
Message-ID: <634@quintus.UUCP>
Date: 5 Nov 88 04:11:49 GMT
References: <156@gloom.UUCP> <18931@apple.Apple.COM> <40@sopwith.UUCP> <10471@s.ms.uky.edu> <611@quintus.UUCP> <1988Nov3.185535.28850@utzoo.uucp>
Sender: news@quintus.UUCP
Reply-To: ok@quintus.UUCP (Richard A. O'Keefe)
Organization: Quintus Computer Systems, Inc.
Lines: 42

In article <1988Nov3.185535.28850@utzoo.uucp> henry@utzoo.uucp (Henry Spencer) writes:
>In article <611@quintus.UUCP> ok@quintus.UUCP (Richard A. O'Keefe) writes:
>>But it _has_ been tossed out of many designs:  the IBM RT PC (at any
>>rate the old ones) had no floating-point instructions, but used a
>>National chip (32081?).

>I.e., it had floating-point hardware.  Why do you say that floating
>point was "tossed out" of its design??

I was responding to someone else who said that floating point ops
_ought_ to be tossed out, and by his phrasing appeared to be implying
that this was a prospect for improvement which RISC designers had so far
ignored.  "tossed out" was _his_ phrase, not mine.

As for whether the RT PC "had floating-point hardware", the floating-
point chip was not accessed as a co-processor, but as a device, and fast
it was not.  By that criterion, an 8008 talking down an RS232 line to a
/370 to get floating-point done could be regarded as "having floating-
point hardware".  The RT's instruction set has _no_ floating-point ops,
not even co-processor ones.

>>The AMD 29000 manual says that floating-point
>>ops are currently extracodes emulated by a trap handler...

>Which may well use the 29xyz (don't remember the number offhand) chip
>to do all the real work.  Again, it hasn't been tossed out, just moved
>around a bit.

Again, your reply misses the point of my message.  I was responding to
someone who claimed that RISC designers _ought_ to be throwing floating-
point calculation out of the CPU chip, and I was merely pointing out
that this had already been done.  (My copy of the AMD29000 manual does
not mention a 29xyz, but says FP is currently done in software.  If you
do recall the part number & what manual to ask for, please let me know.)

>So?  On most of those chips, the software can't tell the difference.
>Does the FP hardware have to be on the CPU chip to be "real" somehow?
>Why?

This is a gross distortion of my position.  You will search my former
posting in vain for any suggestion that offboarding floating point is
_bad_, only that it has already been _done_.