Message-Id: <200001131122.FAA14002@lakdiva.slt.lk>
From: "Kalum Somaratna aka Grendel" <kalum AT newmail DOT net>
To: djgpp AT delorie DOT com
Date: Thu, 13 Jan 2000 05:22:25 +0600
MIME-Version: 1.0
Content-type: text/plain; charset=US-ASCII
Content-transfer-encoding: 7BIT
Subject: Re: amd processors and assembly language
In-reply-to: <387CB637.77A70DC0@slu.edu>
X-mailer: Pegasus Mail for Win32 (v3.12)
Reply-To: djgpp AT delorie DOT com
Errors-To: dj-admin AT delorie DOT com
X-Mailing-List: djgpp AT delorie DOT com
X-Unsubscribes-To: listserv AT delorie DOT com
Precedence: bulk

On 12 Jan 00, at 11:13, John L. Gammel wrote:

> Thanks for the information that assembly language for amd processors
> is more or less identical to the instruction set for the Intel 486
> architecture. The trouble I have encountered is this:  A friend has
> recently downloaded the djgpp package (I have been using mine for more
> than 1 year) and tried to compile some of my source codes.  He has an
> AMD processor.  It all went ok except when compiling my codes using
> assembly language.  That also went ok except when using the -O2
> optimization switch (gxx codename.cc -o codename.exe -O2).  Strangely
> enough, the executable crashed in that case. 

This maybe a compiler bug. Do you Still have the source code by 
any chance. If so could you please post it.

BTW Did this executable run on Non AMD  (Intel) CPU's

> Also, we noted that the
> codes ran (on the machine with an AMD processor) as fast without the
> -O2 switch as we thought they would with the -O2 chip (as though
> optimization was not necessary with the AMD processor).

That may be because the AMD (especially the K7)is a RISC chip. 
It can "optimize/reduce" unoptimized code to optimized version 
before executing.

>      My question is: does anyone have similar experience?  And what
>      are
> the
> optimization switches for the AMD processor?.
> 

BW What is your AMD cpu. Is it a K6 or that super King of the x86 
chip's, the ATHLON (blazingly fast it beats every x86 intel can 
produce, it's speed is like linux vs windoze) .

Anyway I use the -O3 and -fomit-frame-pointer switches which 
enable full optimization. But you might be a bit careful with asking 
the compiler to perform inline optimizations as heavily inlined code 
may overflow the CPU cache and lead to unnecessary slowdowns.

Take Care,
Kalum