From owner-freebsd-current  Tue Aug  6 19:50:59 1996
Return-Path: owner-current
Received: (from root@localhost)
          by freefall.freebsd.org (8.7.5/8.7.3) id TAA25826
          for current-outgoing; Tue, 6 Aug 1996 19:50:59 -0700 (PDT)
Received: from relay.hp.com (relay.hp.com [15.255.152.2])
          by freefall.freebsd.org (8.7.5/8.7.3) with ESMTP id TAA25820
          for <freebsd-current@freebsd.org>; Tue, 6 Aug 1996 19:50:47 -0700 (PDT)
Received: from fakir.india.hp.com by relay.hp.com with ESMTP
	(1.37.109.16/15.5+ECS 3.3) id AA254646237; Tue, 6 Aug 1996 19:50:41 -0700
Received: from localhost by fakir.india.hp.com with SMTP
	(1.37.109.16/15.5+ECS 3.3) id AA276586370; Wed, 7 Aug 1996 08:22:50 +0530
Message-Id: <199608070252.AA276586370@fakir.india.hp.com>
To: "Michael L. VanLoon -- HeadCandy.com" <michaelv@HeadCandy.com>
Cc: joerg_wunsch@uriah.heep.sax.de (Joerg Wunsch),
        freebsd-current@freebsd.org (FreeBSD-current users)
Subject: Re: Patch to recognize AMD 5x86 cpus 
In-Reply-To: Your message of "Tue, 06 Aug 1996 08:47:33 MST."
             <199608061547.IAA18751@MindBender.HeadCandy.com> 
Date: Wed, 07 Aug 1996 08:22:49 +0530
From: A JOSEPH KOSHY <koshy@india.hp.com>
Sender: owner-current@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk

>>>> ""Michael L. VanLoon -- HeadCandy.com"" <michaelv@HeadCandy.com> writes
> 
> >As Michael L. VanLoon -- HeadCandy.com wrote:
> >> If you wanted to do this The "Correct" Way, you might be better off
> >> rewriting the CPU identification code in locore.s to a more general,
> 
> >                                           identcpu.c
> >(It has been divorced from locore.s some time ago.)
> 
> Is it modular, well abstracted, comprehensive and all encompassing?  I

Well from my (possibly inadequate) understanding, we now have a new
instruction `CPUID', which returns information about the CPU in the
register.  Sub-functions are requested using a sub-function code 
specified in %eax.

In order to detect if the CPUID instruction exists, the documented approach
is to check the writability of a bit in EFLAGS.  If the instruction exists,
we call it with %eax = 0 and 1 to extract the standard information about 
the chip.

>From this point the CPU implementations are allowed to diverge.  AMD for 
example, specifies that their K5 series of chips return more information 
about internal organization etc when %eax = 0x80000000-0x80000005.  Other 
vendors support different subfunction codes.

The code in -current retrieves the `basic' CPUID information in locore.s
and updates some globals.  identcpu.c then looks at this information to
determine the exact type/model of CPU and takes the appropriate action
specific to the chip.

To the best of my understanding, the -current code confirms to the `standard' 
intel recommended processor identification method and my code confirms 
to AMD's published processor identification procedure.

For more information you could look at intels processor recognition app-note
in the developers corner of their web-site.  AMD's documents on this
are similarly available on the web.

Koshy
<koshy@india.hp.com>		My Personal Opinions Only