Date: Sun, 21 Dec 1997 14:29:05 +0200 (IST)
From: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
To: DJ Delorie <dj AT delorie DOT com>
cc: djgpp-workers AT delorie DOT com
Subject: Re: Alternate malloc?
In-Reply-To: <199711281928.OAA03216@delorie.com>
Message-ID: <Pine.SUN.3.91.971221142315.8399d-100000@is>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Precedence: bulk


On Fri, 28 Nov 1997, DJ Delorie wrote:

> I've published my results online:
> 
> 	http://www.delorie.com/djgpp/malloc/
> 
> Discussion and comments welcome.

I've tested this on DOS and Windows 95 and got results which are very
similar to DJ's graphs (except for the timing, which was close, but not
identical: it was on a P166).  Results (*.out and *.rout files) available
upon request.  I couldn't test the native SGI versions and neither their 
libmalloc version, but everything else was tested, including GNU malloc 
(I got it from the latest Emacs distribution).

Btw, test2 and test3 bomb with CWSDPMI (they work with QDPMI); seems like 
a NULL ptr dereference, but I didn't have time to look into it.  
Tracebacks available upon request.

Seems like the later versions (4 or 6) are good candidates to be included 
in v2.02.  How about making test6 the default malloc and making the BSD 
version available in a separate library?