Subject: Re: improving ssh performance on sun4m systems
To: None <netbsd-users@netbsd.org>
From: Charles Shannon Hendrix <shannon@widomaker.com>
List: netbsd-users
Date: 03/11/2002 13:30:24
On Mon, Mar 11, 2002 at 04:38:30PM +0000, David Brownlee wrote:

> > I installed that package, and on an SS5/110 microSPARC machine,
> > this is what it says:
> >
> > -mcpu=supersparc
> >
> > This machine is not a supersparc machine.
> 
> 	My athlong is not a pentium pro machine, but -march=pentiumpro is
> 	the best option that gcc has. If you know of a better cpu option
> 	than supersparc for microSPARC I'd be very happy to update
> 	cpuflags to match.

The Athlon is a later generaltion clone of the Intel, and even
borrowed some ideas from the PPro.  A better analogy might be running
-march=pentiumpro on a Pentium or 486, which can slow performance or
even break code.

Anyway, my comment was incomplete: I meant to ask if the supersparc
differed in instruction set at all, and would the optimizations
be good for a microsparc.

So far the changes in code I've looked at are only in instruction
ordering.

Maybe the names of the CPUs are basically meaningless in this context,
and microSPARC just lacks cache and/or other internal optimizations.


-- 
UNIX/Perl/C/Pizza__________________________________shannon@widomaker.com