From broadley@neurocog.lrdc.pitt.edu Mon Jun 28 12:29:32 1993
Return-Path: <broadley@neurocog.lrdc.pitt.edu>
Received: from gossip.pyramid.com by dutecai.et.tudelft.nl (4.1/1.34JP)
          id AA25068; Mon, 28 Jun 93 12:29:29 +0200
Received: from sword.eng.pyramid.com 
	by gossip.pyramid.com (5.61/OSx5.1a Pyramid-Internet-Gateway)
	id AA09784; Mon, 28 Jun 93 03:29:24 -0700
Received: from goss.pyramid.com
	by sword.eng.pyramid.com (5.61/Pyramid_Internal_Configuration)
	id AA09072; Mon, 28 Jun 93 03:29:05 -0700
Received: from neurocog.lrdc.pitt.edu 
	by gossip.pyramid.com (5.61/OSx5.1a Pyramid-Internet-Gateway)
	id AA09759; Mon, 28 Jun 93 03:29:03 -0700
Message-Id: <9306281029.AA09759@gossip.pyramid.com>
Received: by neurocog.lrdc.pitt.edu
	(1.37.109.4/16.2) id AA20804; Mon, 28 Jun 93 06:28:32 -0400
From: Bill Broadley <broadley@neurocog.lrdc.pitt.edu>
Subject: Proposed to comp.arch
To: riscy@pyramid.com (Mips 3000)
Date: Mon, 28 Jun 1993 06:28:32 -0500 (EDT)
X-Mailer: ELM [version 2.4 PL21]
Content-Type: text
Content-Length: 1451      
Status: RO
X-Status: 

I was thinking of posting this with 2 goals: Explore any other options for
cpu, and get more people interested in design, and purchase of the board.
------------------------------------------------------------------------
Subject: Best cpu under $300

I'm looking for the best cpu to put in a motherboard to be designed for a 
small quantity production run (say between 25-100) to run a free unix port.  

Mandatory considerations: 
   Maximum to memory cpu speed 40 Mhz (for ease of design)
   Must run gcc (for easy of porting a free unix).
   Around $300 or less in quantity of 25-100  (going for a < $700 motherboard)

How do I define best?  
	Maximizing speed (specint, specfp) 
	minimizing cost/complexity of the implementation.  
	Cleaness of design/architecture.
	Ease of programming.

Optional floating point would be great for those who can't afford it.
 
So this probably rules out any Mips 4000, Alpha, Pentium chips on cost alone.

I've heard that the i860 and i960 are tough to program, and I don't like the 
design of the 486.

Post or email.
-----------------------------------------------------------------------
Optionally:

If you interested in the design of the motherboard send a subscribe to 
riscy-request@pyramid.com

-- 
Bill					1st>	Broadley@neurocog.lrdc.pitt.edu
Broadley@schneider3.lrdc.pitt.edu <2nd 	3rd> 	             Broadley+@pitt.edu
Linux is great.         Bike to live, live to bike.                      PGP-ok


 
From broadley@neurocog.lrdc.pitt.edu Mon Jun 28 14:56:59 1993
Return-Path: <broadley@neurocog.lrdc.pitt.edu>
Received: from liberator.et.tudelft.nl by dutecai.et.tudelft.nl (4.1/1.34JP)
          id AA25522; Mon, 28 Jun 93 14:56:56 +0200
Received: from neurocog.lrdc.pitt.edu by liberator.et.tudelft.nl (4.1/1.34JP)
          id AA05251; Mon, 28 Jun 93 14:56:50 +0200
Message-Id: <9306281256.AA05251@liberator.et.tudelft.nl>
Received: by neurocog.lrdc.pitt.edu
	(1.37.109.4/16.2) id AA21160; Mon, 28 Jun 93 08:56:14 -0400
From: Bill Broadley <broadley@neurocog.lrdc.pitt.edu>
Subject: Re: Proposed to comp.arch
To: wolff@dutecai.et.tudelft.nl (Rogier Wolff)
Date: Mon, 28 Jun 1993 08:56:14 -0500 (EDT)
In-Reply-To: <9306281241.AA05161@liberator.et.tudelft.nl> from "Rogier Wolff" at Jun 28, 93 02:41:04 pm
X-Mailer: ELM [version 2.4 PL21]
Content-Type: text
Content-Length: 804       
Status: RO
X-Status: 

> 
> Your post to Comp.arch mentioned the I960 as "hard to program".
> This is not true. The 960 has a clean architecture, and has a GCC 
> port.
> 
> 				Roger.

Hmmm I heard several things that make it unattractive for programming
in unix.  (which I kinda believe because nobody uses it for unix).

Does it do context switching well?  I have heard things like mult a,b,c
Multiplies a*b and puts the answer from 3 multiplies ago in c.

It's supposed to be very fast (at least in fp) but very hard to
extract that performance.  Any info would be appreciated.

There's an i860 port for gcc also.

-- 
Bill					1st>	Broadley@neurocog.lrdc.pitt.edu
Broadley@schneider3.lrdc.pitt.edu <2nd 	3rd> 	             Broadley+@pitt.edu
Linux is great.         Bike to live, live to bike.                      PGP-ok

 
From broadley@neurocog.lrdc.pitt.edu Mon Jun 28 15:55:06 1993
Return-Path: <broadley@neurocog.lrdc.pitt.edu>
Received: from liberator.et.tudelft.nl by dutecai.et.tudelft.nl (4.1/1.34JP)
          id AA25792; Mon, 28 Jun 93 15:55:03 +0200
Received: from neurocog.lrdc.pitt.edu by liberator.et.tudelft.nl (4.1/1.34JP)
          id AA05586; Mon, 28 Jun 93 15:54:58 +0200
Message-Id: <9306281354.AA05586@liberator.et.tudelft.nl>
Received: by neurocog.lrdc.pitt.edu
	(1.37.109.4/16.2) id AA21329; Mon, 28 Jun 93 09:54:27 -0400
From: Bill Broadley <broadley@neurocog.lrdc.pitt.edu>
Subject: Re: Proposed to comp.arch
To: wolff@dutecai.et.tudelft.nl (Rogier Wolff)
Date: Mon, 28 Jun 1993 09:54:27 -0500 (EDT)
In-Reply-To: <9306281332.AA05485@liberator.et.tudelft.nl> from "Rogier Wolff" at Jun 28, 93 03:32:11 pm
X-Mailer: ELM [version 2.4 PL21]
Content-Type: text
Content-Length: 800       
Status: RO
X-Status: 

> 
> The pipeline in software (mul a,b,c -> result from two multiplies back in
> c) is an '860 "feature".
Ahh I guess I got confused...

> 
> The '960 is inattractive for a Unix implementation. In that you are right.
> 
Well thats definitely the target, I change the wording thanks.

> The different versions of the '960 have either a MMU, FP or a Super Scalar
> version. No two of these go together (I think... I use the Super Scalar
> version: the I960 CA). For Unix you need all three. (The military
> is able to get two of them.....)
> 
I'd guess FP and Superscaler.  No need for MMU.


-- 
Bill					1st>	Broadley@neurocog.lrdc.pitt.edu
Broadley@schneider3.lrdc.pitt.edu <2nd 	3rd> 	             Broadley+@pitt.edu
Linux is great.         Bike to live, live to bike.                      PGP-ok


 
