X-pop3-spooler:	POP3MAIL 2.1.0 b 4 980420 -bs-
Message-ID:	<19980902171402.50556@cerebro.laendle>
Date:	Wed, 2 Sep 1998 17:14:02 +0200
From:	Marc Lehmann <pcg AT goof DOT com>
To:	Krzysztof Strasburger <strasbur AT chkw386 DOT ch DOT pwr DOT wroc DOT pl>
Cc:	beastium <beastium-list AT Desk DOT nl>
Subject:	Re: pgcc-1.1a - first impression
Mail-Followup-To:	Krzysztof Strasburger <strasbur AT chkw386 DOT ch DOT pwr DOT wroc DOT pl>,
	beastium <beastium-list AT desk DOT nl>
References:	<m0zEBKp-0000FVC AT chkw386 DOT ch DOT pwr DOT wroc DOT pl>
Mime-Version:	1.0
In-Reply-To:	<m0zEBKp-0000FVC@chkw386.ch.pwr.wroc.pl>; from Krzysztof Strasburger on Wed, Sep 02, 1998 at 11:44:00AM +0000
X-Operating-System:	Linux version 2.1.119 (root AT cerebro) (gcc version pgcc-2.91.54 19980816 (gcc2 ss-980609 experimental))
Status:	RO
Lines:	48

On Wed, Sep 02, 1998 at 11:44:00AM +0000, Krzysztof Strasburger wrote:
> Hi!
> I just downloaded and compiled pgcc-1.1a and did a simple comparison
> with the older pgcc-1.0.3a (my own FPU-intensive program).
> Compilation options:
> -march=pentium -mcpu=pentium -On -ffast-math -frerun-loop-opt
> -malign-double -mstack-align-double -malign-jumps=0 -malign-loops=0
> -malign-functions=0,
> where n=2,3,4 or s
> The first thing I noticed is code bloat :-(, but this is rather egcs

Does -fno-exceptions help?

> pgcc-1.0.3a: -O2    -O3    -O4
>              46600  46920  48024
> pgcc-1.1a: -O2    -O3    -O4    -Os
>            47880  48296  49224  46888
> The second thing I noticed is no speed improvement on Pentium :-(((
> (this is valid for tested program only, of course).

Thats, erhm, bad. But did you try -O6, as the most effective optimization
on pentiums is enabled with -O5 and higher only.

Having said this, I have no idea what could have happened to pentium opts,
as I can't remember changing them significantly ;( But the performance
problems in egcs/pgcc I talked about in my earlier mail is mainly related to
fpu-intensive programs...

Could you try without the *align-double options? (just a guess) And maybe
you could post your benchmark data with -O2,3,6?

> And now the good news. The optimization on Pentium Pro is better than before.

*sigh*

> The code optimized for size performs surprisingly well.

Could be better ;) inlining still bloats code (it shouldn't).

Thanks,

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

- Raw text -