Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 5 Dec 2002 15:26:38 +0200 (WET)
From:      Evren Yurtesen <eyurtese@turkuamk.fi>
To:        David Schultz <dschultz@uclink.Berkeley.EDU>
Cc:        freebsd-questions@FreeBSD.ORG
Subject:   Re: -O3 optimization?
Message-ID:  <Pine.A41.4.10.10212051522150.93290-100000@bessel.tekniikka.turkuamk.fi>
In-Reply-To: <20021205131130.GB11161@HAL9000.homeunix.com>

next in thread | previous in thread | raw e-mail | index | archive | help
Well, I made searches from google and people talk that O3 produced quite
noticably faster code. But well I am not so hungry for speed. I just
wondered if the binary might have something wrong with it or not even
though the compiler didnt complain while compiling.

What about using -O or not using any optimizations? Is it very rare that
-O breaks somethings? I was using -Os and I also didnt notice anything
wrong but maybe there can be something I am missing too...

Is there big performance improvement between -O and -O2 ? or from not
using any optimizations to -O or -O2? Lets say if I am compiling
KDE,XFree86. How much would it effect? is there a web page with some
statistical data about this?

Evren

On Thu, 5 Dec 2002, David Schultz wrote:

> Thus spake Evren Yurtesen <eyurtese@turkuamk.fi>:
> > I wonder if a source is compiled with -O3 without any problems, might
> > there be any problems in binaries which might create crashes?
> 
> You're welcome to try it out, but it isn't supported.  GCC has a
> few obscure misfeatures at -O3.  Some applications break at -O3,
> usually because they violate C's aliasing rules or contain broken
> inline assembly.  At one point, the kernel's TCP checksum code had
> some difficult-to-solve problems with -O3, and I'm not sure
> whether that has been fixed.
> 
> Despite all of that, I built world and kernel with -O2 a while ago
> and noticed no problems whatsoever.  Just note that you've been
> warned, and you probably won't see a significant performance
> improvement anyway.
> 


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-questions" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.A41.4.10.10212051522150.93290-100000>