Path: utzoo!utgpu!news-server.csri.toronto.edu!clyde.concordia.ca!uunet!samsung!usc!apple!sun-barr!newstop!sun!amdcad!weitek!practic!vlsisj!davidc
From: davidc@vlsisj.VLSI.COM (David Chapman)
Newsgroups: comp.object
Subject: Re: code blocks (aka closures)
Message-ID: <15625@vlsisj.VLSI.COM>
Date: 6 Jul 90 00:54:12 GMT
References: <12396@june.cs.washington.edu> <1112@carol.fwi.uva.nl> <5319@stpstn.UUCP> <11013@alice.UUCP>
Reply-To: davidc@vlsisj.UUCP (David Chapman)
Organization: VLSI Technology Inc., San Jose, CA
Lines: 41

In article <11013@alice.UUCP> bs@alice.UUCP (Bjarne Stroustrup) writes:
>Software is not hardware. Further, there is little reason to believe
>that software is analogous to hardware in any strong and useful sense;
>had it been, we probably wouldn't have had a `software crisis.' Looking
>to hardware for solutions to software problems (except by using better
>hardware) is likely to be highly misleading and causes much confusion.

I believe exactly the opposite:  that software is exactly analogous to
hardware.  It's just that the software is one or two orders of magnitude
more complex at any given point in time.

We write software to help system designers build ICs.  If it weren't for
CAD software from us, other companies, and universities, there wouldn't
be enough chip designers to go around.

I do agree that we cannot look to hardware for solutions to software
problems.  If anything it's the other way around.  We use software to
build hardware.  Eventually you may be able to write code and have your
CAD software automatically generate a chip for you that executes the
code directly.

>The problems with software are fundamental. They are primarily caused by
>increasing ambition on the part of individuals and organizations: As long
>as we are seeing rapid progress many of us will be working at the hairy
>edge of our abilities and of the capabilities of our tools.

We see exactly this problem with chip designers.  There's just an extra limit:
die size (corresponding to number of lines of code).  It can't be too big
or else it is physically unmanufacturable.  The Motorola 68000, with about 
68,000 transistors, is now a small design.  We are now building half-million 
transistor custom ICs.

And I find I design hardware the same way I write software.  A NAND gate is
in some sense the Turing machine of logic designers.  After all, our software
runs on hardware.  Why shouldn't they be similar at the most fundamental
levels?
-- 
		David Chapman

{known world}!decwrl!vlsisj!fndry!davidc
vlsisj!fndry!davidc@decwrl.dec.com