Date: Sun, 7 May 2000 11:21:26 +0300 (IDT)
From: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
X-Sender: eliz AT is
To: Laurynas Biveinis <lauras AT softhome DOT net>
cc: djgpp-workers AT delorie DOT com
Subject: Re: Perfomance of gc-simple
In-Reply-To: <3914F929.A9577BB8@softhome.net>
Message-ID: <Pine.SUN.3.91.1000507111811.21782K-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 Sun, 7 May 2000, Laurynas Biveinis wrote:

> > However, if alignment is the only difference between `valloc' and
> > `malloc', I don't quite see what's the big difference between them
> > that would explain such a huge run-time penalty.  Can someone explain?
> 
> Because they're used in two different GC implementations: gc-simple uses
> malloc and gc-page uses mmap or valloc.

It sounds like you could try to solve this by simply having the right 
#define's in the system headers, to dupe the configure script into 
thinking we have valloc.  I suggest to see how well does this work.

As for mmap, it should be trivial to throw together a KISS[1] 
implementation that simply reads the entire file into memory, and see 
how well (or how badly ;-) does that work, especially with very large 
sources.

[1] KISS = Keep It Simple and Stupid