From owner-freebsd-questions  Mon Sep 23 06:48:57 1996
Return-Path: owner-questions
Received: (from root@localhost)
          by freefall.freebsd.org (8.7.5/8.7.3) id GAA29749
          for questions-outgoing; Mon, 23 Sep 1996 06:48:57 -0700 (PDT)
Received: from dyson.iquest.net (dyson.iquest.net [198.70.144.127])
          by freefall.freebsd.org (8.7.5/8.7.3) with ESMTP id GAA29724
          for <freebsd-questions@FreeBSD.ORG>; Mon, 23 Sep 1996 06:48:51 -0700 (PDT)
Received: (from root@localhost) by dyson.iquest.net (8.7.5/8.6.9) id IAA10130; Mon, 23 Sep 1996 08:48:37 -0500 (EST)
From: "John S. Dyson" <toor@dyson.iquest.net>
Message-Id: <199609231348.IAA10130@dyson.iquest.net>
Subject: Re: ksh [was: Re: Shells shells shells?]
To: zgabor@CoDe.hu (Gabor Zahemszky)
Date: Mon, 23 Sep 1996 08:48:37 -0500 (EST)
Cc: freebsd-questions@FreeBSD.ORG, tony@rtd.com
In-Reply-To: <199609231018.KAA01709@CoDe.CoDe.hu> from "Gabor Zahemszky" at Sep 23, 96 10:18:41 am
Reply-To: dyson@FreeBSD.ORG
X-Mailer: ELM [version 2.4 PL24 ME8]
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: owner-questions@FreeBSD.ORG
X-Loop: FreeBSD.org
Precedence: bulk

> 
> > Anyway the version I have (downloaded ~ Aug 95) is larger than even bash.
> > 
> > $ size /usr/local/bin/ksh
> > text    data    bss     dec     hex
> > 409600  16384   8724    434708  6a214
> 
I just measured the memory usage of bash vs. ksh93.  It is true that the
.text is larger for the new ksh.  However, the per-process .data/.bss is
much smaller.  Specifically, bash is using approx 68pages!!! of .data/.bss
per invocation.  Ksh uses about 18 pages per invocation.  But bash
uses 69pages of constant (and shared) .text and ksh uses 97 pages
of constant (and shared) .text.  If you are going to choose between
ksh and bash on a heavily used system, it appears that ksh might use
less real memory in the long run.  (Note that the /proc/xxx/map
file is useful for figuring these things out!!!)  It is really
amazing that bash uses so much memory.  I haven't measured the
memory usage when some of the pages have been reclaimed by paging.
Bash is probably using less pages than indicated after process
startup though.

Since shell processes are used so often, it would be interested
to benchmark a system (under load) with the various shells to
find out which ones function best under loaded ISP or server-type
conditions.

John