From: Hans-Bernhard Broeker <broeker AT physik DOT rwth-aachen DOT de>
Newsgroups: comp.os.msdos.djgpp
Subject: Re: speed of compiled programms
Date: 4 Sep 2001 19:20:27 GMT
Organization: Aachen University of Technology (RWTH)
Lines: 25
Message-ID: <9n39hr$jr8$1@nets3.rz.RWTH-Aachen.DE>
References: <80o7ptk35260gn4l05obuh0du6k37psstq AT 4ax DOT com> <Pine DOT SUN DOT 3 DOT 91 DOT 1010904111633 DOT 28761F-100000 AT is> <2p1aptcsq4ra5m2vb0mor7553fcbipv50l AT 4ax DOT com>
NNTP-Posting-Host: acp3bf.physik.rwth-aachen.de
X-Trace: nets3.rz.RWTH-Aachen.DE 999631227 20328 137.226.32.75 (4 Sep 2001 19:20:27 GMT)
X-Complaints-To: abuse AT rwth-aachen DOT de
NNTP-Posting-Date: 4 Sep 2001 19:20:27 GMT
Originator: broeker@
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp
Reply-To: djgpp AT delorie DOT com

Marco Vienup <Marco AT vienup DOT de> wrote:
>>How much slower?  Can you give numbers?

> Hm, I have a simple timeout for(...) construct.

> for(int i=0; i<=500;i++){ one small if(...) }
> If it is the fast code, it will run in a timeout, with 
> the slow code the maximum is 100-120.

You could be a bit more precise.  The precise nature of that "one
small if()", and the exact set of compiler switches, along with the
compiler version might well be needed to reproduce the behaviour you
observed.  It may also be helpful if you run gcc with the compiler
switch -save-temps added, and then look at (or show us) the difference
in the *.s files generated that way, for the different optimization levels.

From a more general point of view, -O3 was indeed known to slow down
code rather than speed them up, for several programs, and several
versions of GCC. That's because the main difference between -O3 and
-O2 used to be the -finline optimization switch, which tends to
deteriorate the performance on architectures tuned for optimal
execution of loops such as the x86 platform.
-- 
Hans-Bernhard Broeker (broeker AT physik DOT rwth-aachen DOT de)
Even if all the snow were burnt, ashes would remain.