From:	"Laurynas Biveinis" <lauras AT softhome DOT net>
Date:	Sun, 29 Apr 2001 08:21:04 +0200
To:	djgpp-workers AT delorie DOT com
Cc:	Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
Subject:	Re: gcc-3.0 related patches for DJGPP CVS version
Message-ID:	<20010429082104.A543@lauras.lt>
Mail-Followup-To:	djgpp-workers AT delorie DOT com,
	Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
References:	<3AE96E9B DOT 10268 DOT B1244D AT localhost> <4331-Fri27Apr2001162351+0300-eliz AT is DOT elta DOT co DOT il> <3AEAF6C8 DOT 19289 DOT 14FAA0A AT localhost>
Mime-Version:	1.0
User-Agent:	Mutt/1.3.15i
In-Reply-To:	<3AEAF6C8.19289.14FAA0A@localhost>; from pavenis@lanet.lv on Sat, Apr 28, 2001 at 04:58:48PM +0300
Reply-To:	djgpp-workers AT delorie DOT com
Errors-To:	nobody AT delorie DOT com
X-Mailing-List:	djgpp-workers AT delorie DOT com
X-Unsubscribes-To:	listserv AT delorie DOT com

> > >   -malign-loops=2
> > >   -malign-jumps=2
> > >   -malign-functions=2
> > 
> > Shouldn't we revise the -malign-* switches as well?  I think these
> > switches generate suboptimal code.
> 
> I think the best would be to remove them all. I built current CVS 
> version both with and without -malign* switches (using gcc-3.0 
> 20010426). Size difference (read bloat ...) is very small (<1%). 

I think we should keep these switches, but with alignment increased
to cater p3-p4 and k7, because some GCC defaults might be suboptimal too.
(Because of backwards compatibility of ABI). Of course in this case
these switches should go into specs too. And there are a few new interesting
options, like -m128bit-long-double.

Laurynas

- Raw text -