Mailing-List: contact cygwin-developers-help AT sourceware DOT cygnus DOT com; run by ezmlm
List-Subscribe: <mailto:cygwin-developers-subscribe AT sourceware DOT cygnus DOT com>
List-Archive: <http://sourceware.cygnus.com/ml/cygwin-developers/>
List-Post: <mailto:cygwin-developers AT sourceware DOT cygnus DOT com>
List-Help: <mailto:cygwin-developers-help AT sourceware DOT cygnus DOT com>, <http://sourceware.cygnus.com/ml/#faqs>
Sender: cygwin-developers-owner AT sourceware DOT cygnus DOT com
Delivered-To: mailing list cygwin-developers AT sourceware DOT cygnus DOT com
Date: Wed, 5 Jul 2000 22:26:07 -0400
From: Chris Faylor <cgf AT cygnus DOT com>
To: cygwin-developers AT sourceware DOT cygnus DOT com
Subject: Re: copy-on-write (oh well)
Message-ID: <20000705222607.A18846@cygnus.com>
Reply-To: cygwin-developers AT sourceware DOT cygnus DOT com
Mail-Followup-To: cygwin-developers AT sourceware DOT cygnus DOT com
References: <20000703225357 DOT A25565 AT cygnus DOT com>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
User-Agent: Mutt/1.2i
In-Reply-To: <20000703225357.A25565@cygnus.com>; from cgf on Mon, Jul 03, 2000 at 10:53:57PM -0400

On Mon, Jul 03, 2000 at 10:53:57PM -0400, cgf wrote:
>I spent part of my "day off" today hacking together a "copy-on-write"
>implementation for copying the heap after a fork.  I've long wanted to
>try to do something like this.  It seems to work pretty well as long as
>I don't fork something that forks something else.
>
>I can't tell if there is a performance improvement or degradation or not
>yet.
>
>I hope that nobody is standing while reading this because I have some
>shocking news.
>
>The copy-on-write functionality is badly broken on Windows 95, so I
>can't use it there.  This only works on NT.

I couldn't figure out how to get copy-on-write working beyond the
first level of fork.  There doesn't seem to be any way to share
a pool of memory inherited from a parent process and potentially
updated with a copy-on-write operation.

It turns out not to matter anyway.  The current version of fork is
somehow, inexplicably faster than my copy-on-write version.

I also tried having the child process inherit the parent's memory
and just memcpy'ing the parent's memory into the child's malloc
pool.  That was the slowest of all.

Oh well.

cgf