Message-ID: <20000518032721.73423.qmail@hotmail.com>
X-Originating-IP: [200.45.92.173]
From: "Norberto Alfredo Bensa" <nbensa AT hotmail DOT com>
To: <djgpp AT delorie DOT com>
References: <Pine DOT SUN DOT 3 DOT 91 DOT 1000517152259 DOT 626E-100000 AT is>
Subject: Re: rebuilding libc...
Date: Thu, 18 May 2000 00:18:24 -0300
Organization: nBens@ Computers
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Priority: 3
X-MSMail-Priority: Normal
X-Mailer: Microsoft Outlook Express 5.50.4029.2901
X-MimeOLE: Produced By Microsoft MimeOLE V5.50.4029.2901
Reply-To: djgpp AT delorie DOT com

Thanks again...

yes, I've tried -mpentiumpro (-mcpu and -march) and alignment of 4... but it
seams that it's even slower...
I've tried -O6 also, but then libm (?) did not finished compiling... I can't
remember the error message... something in dbgcom.c (I can compile it again
if you want...)

btw, I have binutils 2.9.5.1... I've downloaded "everything" last weekend...

Greetings, and thanks so much for your help,
Norberto



----- Original Message -----
From: "Eli Zaretskii" <eliz AT is DOT elta DOT co DOT il>
To: "Norberto Alfredo Bensa" <nbensa AT hotmail DOT com>
Cc: <djgpp AT delorie DOT com>
Sent: Wednesday, May 17, 2000 9:29 AM
Subject: Re: rebuilding libc...


>
> On Wed, 17 May 2000, Norberto Alfredo Bensa wrote:
>
> > > > o    What are the best compiler options when rebuilding the standard
> > > > libraries if I'm using a pentium II?
> > >
> > > It depends on the version of Binutils that you have installed.  What
> > > does "as --version" print?
> > >
> > here it is. I compiled it my self....
> >
> > GNU assembler 2.9.5
>
> Then I think you should use the original gcc.opt from djlsr203.zip.  With
> Binutils 2.9.5.1, available from DJGPP sites, you could try higher
> alignment options (-malign-loops=4 -malign-loops=4 -malign-jumps=4), but
> I don't think your version of Binutils will support this correctly.
>
> You could play with -mpentium instead of -m486, but I doubt if that will
> yield significant differences.
>
> Anyway, it only makes sense to try to rebuild the library with more
> optimizations if you have evidence that some library functions consume a
> significant portion of the run time.  The profiler should show this.
> Even if some function does need to be optimized, it is better to compile
> only that function with different optimization switches.
>