Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!husc6!cmcl2!beta!hc!ames!amelia!msf
From: msf@amelia (Michael S. Fischbein)
Newsgroups: comp.misc
Subject: Re: Faster/cheaper execution of Unix pipelines?  A proposal.
Message-ID: <3126@ames.arpa>
Date: Fri, 16-Oct-87 08:01:14 EDT
Article-I.D.: ames.3126
Posted: Fri Oct 16 08:01:14 1987
Date-Received: Sat, 17-Oct-87 18:35:13 EDT
References: <385@ohlone.UUCP>
Sender: usenet@ames.arpa
Reply-To: msf@amelia.UUCP (Michael S. Fischbein)
Organization: NASA Langley Research Center, Hampton, VA
Lines: 46
Keywords: Unix pipes, program composition, small is beautiful

In article <385@ohlone.UUCP> nelson@ohlone.UUCP (Bron Nelson) writes:
>There has been considerable discussion (mostly in comp.arch) about
>"Big Programs Hurt Performance" and the relative merits of a single
>program with lots of options, vs. communicating small programs,
>in particular the use of pipes in the Unix world.  The canonical
>example used is the Unix "ls" command producing single-column and
>multi-column output, as opposed to piping single column output
>through "pr."
> I think the answer is not to give up on "small is beautiful;"
> I ask for your comments, proposals, and
>examples of existing systems that do a good job.
>
>I'll use the simple pipeline "ls | pr -4" as my example.

OK, so will I.

>  I'll call this mythical beast the "Pipeline Composer,"

Goes on to describe a way of implementing the Pipeline Composer.


Bron works for Cray and probably has their machines in mind when he
conceptualizes a program; his technique is intended to minimize the
number of (high overhead) fork/execs, among other things.  But on
most machines, a fork/exec does not impose the huge overhead it does
on a Cray (well, I don't know abou XMPs or 1's, but the overhead on
the 2 is awfully big compared to other UNIX boxes).  One major complaint
about typing
	ls | pr -4 -t  instead of
	ls -C	(the -C is needed in UNICOS)
is the extra typing involved and the additional requirements it places
on the (perhaps naive) user.  One major complaint about the list of
options approach is the size and complexity of each utility.

A reasonable compromise (and one that I think is taken by many of the
utilities) is to use the "system" or "popen" system calls available.

Ls doesn't have to develop and debug a column routine; a system
call to pr will work.  That doesn't involve creating a new Pipeline
Composer, either.
		mike

Michael Fischbein                 msf@prandtl.nas.nasa.gov
                                  ...!seismo!decuac!csmunix!icase!msf
These are my opinions and not necessarily official views of any
organization.