Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 23 May 2000 21:52:25 +0200
From:      Martin Cracauer <cracauer@cons.org>
To:        Doug Barton <Doug@gorean.org>
Cc:        Martin Cracauer <cracauer@cons.org>, "Andrey A. Chernov" <ache@FreeBSD.ORG>, Kris Kennaway <kris@FreeBSD.ORG>, cvs-committers@FreeBSD.ORG, cvs-all@FreeBSD.ORG
Subject:   Re: VM load with static binaries (Re: cvs commit: ports/shells/bash2)
Message-ID:  <20000523215225.A17437@cons.org>
In-Reply-To: <Pine.BSF.4.21.0005231222000.47409-100000@dt051n0b.san.rr.com>; from Doug@gorean.org on Tue, May 23, 2000 at 12:27:31PM -0700
References:  <20000523192328.A16012@cons.org> <Pine.BSF.4.21.0005231222000.47409-100000@dt051n0b.san.rr.com>

next in thread | previous in thread | raw e-mail | index | archive | help
In <Pine.BSF.4.21.0005231222000.47409-100000@dt051n0b.san.rr.com>, Doug Barton wrote: 
> So, I
> would think that since a lot of stuff is already using those two
> libraries, (most significantly, xterm, which is where almost all of my
> bash's run) that it would use less system resources to have the shell
> linked dynamically. 

The shell gets significantly slower and it slows the whole system
down.  See the little benchmark I appended to an earlier mail in this
thread, especially take into account what amount of user and system
CPU time is in use and what kind of operation the system time in the
dynamic linking case consists of.  VM mapping operations.  Then take
another breath and think over the effects in the SMP case.

As you noted, it links only to two librarie and the static linker
takes only those object files from the *.a it needs.

So, the win of static is significant and the cost is very small, both
in disk space and in VM space.

> This is something that I'd be happy to test 
> empirically if you want to show me precise steps to do so. I'm actually
> eager to learn more about this.

As I said, you should look up the mail archives for the postings of
John Dyson regarding this issue.  He's much better to explain the
connections and effects and - more important - he can rate the effects
for the whole system when it is under load, better than any benchmark
we build from scratch can.

Martin
-- 
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Martin Cracauer <cracauer@cons.org> http://www.cons.org/cracauer/
BSD User Group Hamburg, Germany     http://www.bsdhh.org/


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe cvs-all" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20000523215225.A17437>