Subject: Re: should we be using -m486?
To: None <alexw+@andrew.cmu.edu, port-i386@NetBSD.ORG>
From: Max Bell <mbell@europa.com>
List: port-i386
Date: 12/14/1995 12:28:00
>From: "Alex R.N. Wetmore" <alexw+@andrew.cmu.edu>

>Excerpts from internet.computing.netbsd.port-i386: 13-Dec-95 Re: should
>we be using -m486?  by M. HeadCandy.com@HeadCan 

>> my 486.  It shouldn't appreciably slow down 386s.  And, the most
>> important thing, it doesn't break on 386s.  People can use these
>> binaries successfully on their 386s until they decide to rebuild their
>> own binaries (if they even care).

>Are you sure it doesn't break on 386s?  At one point AFS for NetBSD
>was being compiled with -m486 and it didn't run on any of my friends
>386s until the AFS/NetBSD developer built a binary compiled without
>-m486.  I thought it might use the extra one or two instructions in the
>486 instruction set...

It would seem rather obvious to me that using the -m486 flag will allow
the compiler to include 486-specific instructions in the resulting code.
If a given program still works on a 386 it is due to luck, and should
not be depended on.

Since the 386 systems are the slowest of the x86 machines, requiring them
to recompile in order to be sure the code will work would be rather rude.

JMTCW.

Max