Date: Sun, 30 Jan 2000 21:11:58 +0100
From: Marc Lehmann <marc AT gimp DOT org>
To: pgcc AT delorie DOT com
Subject: Re: pgcc and egcs alignment -- function, basic block and string
Message-ID: <20000130211158.D641@cerebro.laendle>
Mail-Followup-To: pgcc AT delorie DOT com
References: <38921CD6 DOT 2A725779 AT ix DOT netcom DOT com> <20000129032101 DOT A25630 AT atrey DOT karlin DOT mff DOT cuni DOT cz> <38940264 DOT 30C35BDD AT neuss DOT netsurf DOT de>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
In-Reply-To: <38940264.30C35BDD@neuss.netsurf.de>; from w.formann@netsurf213.neuss.netsurf.de on Sun, Jan 30, 2000 at 10:20:36AM +0100
X-Operating-System: Linux version 2.2.14 (root AT cerebro) (gcc version 2.95.1 19990816 (release)) 
Reply-To: pgcc AT delorie DOT com
Errors-To: dj-admin AT delorie DOT com
X-Mailing-List: pgcc AT delorie DOT com
X-Unsubscribes-To: listserv AT delorie DOT com
Precedence: bulk

On Sun, Jan 30, 2000 at 10:20:36AM +0100, Wolfgang Formann <w DOT formann AT netsurf213 DOT neuss DOT netsurf DOT de> wrote:
> For whatever reason, I added three nop's in from of the assembler stuff
> in libdes's implementation of fcrypt, I got a speedup of 10%.
> The funny thing is, that now the main loop does start ad some odd
> address.

I did the same in my zip-cracker ;) 3% speedup (and no jump anywhere near
that place ;)

> Well, my processor is an AMD-K6-I/233, so maybe another issue for
> aligning could be to prevent multi-byte instructions not to go over some
> cache (or whatever) boundary, which is the only explanation I curently
> have.

The pentium suffers from this a lot, but the individual speed improvement
is far from your 10%.

> 10% is really a lot, inside a loop, which takes (about) 25 * 35 cycles.

That's very much. I doubt it really is the three nops, but...

-- 
      -----==-                                             |
      ----==-- _                                           |
      ---==---(_)__  __ ____  __       Marc Lehmann      +--
      --==---/ / _ \/ // /\ \/ /       pcg AT opengroup DOT org |e|
      -=====/_/_//_/\_,_/ /_/\_\       XX11-RIPE         --+
    The choice of a GNU generation                       |
                                                         |