Date: Tue, 9 May 2000 11:54:19 +0300 (IDT)
From: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
X-Sender: eliz AT is
To: Martin Str|mberg <ams AT ludd DOT luth DOT se>
cc: djgpp-workers AT delorie DOT com
Subject: Re: Perfomance of gc-simple
In-Reply-To: <200005082100.XAA11742@father.ludd.luth.se>
Message-ID: <Pine.SUN.3.91.1000509115403.946G-100000@is>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
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
Precedence: bulk


On Mon, 8 May 2000, Martin Str|mberg wrote:

> > In principle, yes.  However, before we put mmap into the library, I think 
> > Someone (tm) should look at several GNU packages which use mmap and 
> > verify that our minimal implementation will do what those packages 
> > expect.  Since the availability of mmap is checked by the configure 
> > script by linking a program, and since GNU packages I've seen always 
> > prefer to use mmap if it is found, if we put mmap into the library for 
> > GCC, we had better made sure it doesn't bite us elsewhere.
> 
> Ok, I have hacked together the mmap() function. Do you have any mmap
> intensive code to suggest I try it out on?

The first would be GCC itself, obviously, since that's the motivation
for doing it.

Grepping through GNU source directories I have on this machine, I see
that the following packages use mmap:

     Binutils
     Grep
     Sed

(Many others also probe the library for mmap, but they use it in the
routines that support gettext/NLS, so our ports mostly don't have that
code linked in.)

I'd suggest trying with Grep first (after GCC).

In any case, if it turns out that this implementation is only good
enough for GCC, we can always call it __lame_mmap or something and
have some #define in GCC sources that will use that.