Message-Id: <3.0.6.32.19990224205612.008be9e0@pop.globalserve.net>
X-Sender: derbyshire AT pop DOT globalserve DOT net (Unverified)
X-Mailer: QUALCOMM Windows Eudora Light Version 3.0.6 (32)
Date: Wed, 24 Feb 1999 20:56:12 -0500
To: pgcc AT delorie DOT com
From: Paul Derbyshire <pderbysh AT usa DOT net>
Subject: Re: list info
Cc: johnny AT entity DOT netcologne DOT de
In-Reply-To: <199902241423.JAA29290@envy.delorie.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Reply-To: pgcc AT delorie DOT com

At 09:23 AM 2/24/99 -0500, you wrote:
>double foo (int i, double d) {
>  int j;
>  for (j =3D 20; j; --j) {
>    i *=3D i;
>    d *=3D d;
>  }
>  return d*(double)i;
>}

j =3D 20??? That looks ill-formed to me. -funroll-all-loops or no
-funroll-all-loops, it should be produced a "parse error before 20". What
were you trying to do ther anyways, assign a hex value? Should be
j = 0x3d20
then. Or did you mistype when writing the email? :-)

>    Option:      Output:
>    -march=3Di386: .*.*.*.*.*.*.*.*.*.*
>    -march=3Di486: .*.*.*.*.*.*.*.*.*.*
>    -march=3Di586: ....*.*.**.*.**.*.**
>    -march=3Di686: ******.*...*...*...*
>    -march=3Dk6  : *..*.*.*.*.*.*.*.*.*

-march=3Di386 is not an architecture flag I ever heard of. Are you sure you
don't mean -march=i386?

>Are the other chip specific optimizers better than the pentium's, or is
>this code really faster on pentium?

If it is I'd be very surprised since it pushes stuff on the stack!


-- 
   .*.  "Clouds are not spheres, mountains are not cones, coastlines are not
-()  <  circles, and bark is not smooth, nor does lightning travel in a
   `*'  straight line."    -------------------------------------------------
        -- B. Mandelbrot  |http://surf.to/pgd.net
_____________________ ____|________     Paul Derbyshire     pderbysh AT usa DOT net
Programmer & Humanist|ICQ: 10423848|