Path: utzoo!utgpu!news-server.csri.toronto.edu!utcs.toronto.edu!cks
Newsgroups: comp.sys.dec
From: cks@hawkwind.utcs.toronto.edu (Chris Siebenmann)
Subject: Re: DCL and EDT for Unix?
Message-ID: <1991Jun21.232838.15076@jarvis.csri.toronto.edu>
Organization: Ziebmef home away from home
References: <1991Jun19.122116.8961@cs.utk.edu> <1991Jun19.205450.26120@convex.com> <1991Jun20.182101.3145@cs.utk.edu> <1991Jun20.212124.22288@convex.com> <13829@mentor.cc.purdue.edu>
Date: 22 Jun 91 03:28:38 GMT
Lines: 33

asg@sage.cc.purdue.edu (The Grand Master) writes:
[Disputes the contention that the extra overhead of doing VMS-like stuff
 in a DCL emulation on Unix is negligible, I think.]

 Beware of the beartrap of shell performance, lest it close on you.
Does anyone have real numbers on the "overhead" of various shells, and
how much they differ? I suspect that it's hard for an interactive shell
to wrack up lots of CPU time because of inherent features, and even
somewhat for a shell running a shell script to do so. Compared to
exec()ing a program, fiddling strings is very fast.

| Now I am even more flabergasted.
| You want users to run DCL on UNICOS????!!!??!?!?!?!
| That is a shame when Cray did such a nice job on UNICOS.
| Especially when you are running on such a high performance machine, you
| should use what has been written to specifically work well with the
| underlying OS.

 People who want raw frothing manic performance from Crays usually
run them with batched applications. Once you've made the decision to
sacrifice some CPU power for user friendliness, you might as well
sacrifice a bit more for VMS user friendliness if you think it's
really important. I am also suspicious of the idea that Cray did
anything in particular to tune their shells to the underlying machine
beyond compiling them with the UNICOS C compiler (and fixing the
inevitable portability problems).

--
    ported program:	  a program which takes constant or increasing
			  effort to port to each new machine.
    portable program: a program which takes less effort to port to
		      each new machine.
cks@hawkwind.utcs.toronto.edu	           ...!{utgpu,utzoo,watmath}!utgpu!cks