Path: utzoo!attcan!uunet!cs.utexas.edu!tut.cis.ohio-state.edu!purdue!gatech!prism!vsserv!stat!stat.fsu.edu!mccalpin
From: mccalpin@masig3.masig3.ocean.fsu.edu (John D. McCalpin)
Newsgroups: comp.arch
Subject: Re: parallel systems
Message-ID: <MCCALPIN.89Oct22095056@masig3.masig3.ocean.fsu.edu>
Date: 22 Oct 89 13:50:56 GMT
References: <10164@encore.Encore.COM> <1989Oct20.165407.5887@mentor.com>
	<1989Oct21.234930.905@world.std.com>
Sender: news@stat.fsu.edu
Organization: Supercomputer Computations Research Institute
Lines: 63

In article <1989Oct21.234930.905@world.std.com> bzs@world.std.com
(Barry Shein) writes:
>Why oh why is are people searching around for a philosopher's stone of
>computing architectures. This isn't science, this is aberrant
>psychology.

While there may be some small component of "holy grail"-itis in this
thread, I think that most people are discussing a different problem
here.  This isn't science --- it is technology plus market forces....

>There exists problems which are best run on (pick one or more) MIMD,
>SIMD, scalar, vector-processors, and/or hybrids. There also exist
>problems which don't care what they're run on, at least not much.

I don't think that anyone disputes this.  The question is whether or
not each of these types of architectures can acquire a sufficient
market to be competitive with whatever architecture is selling best,
and which therefore has the most R&D money and the best economies of
scale. 

>To say (MIMD/SIMD) processors are a bad idea because there exists some
>large set of problems which are either impossible or very hard to
>optimize for those architectures is so goddamn stupid it boggles the
>mind.

On the other hand, it is a perfectly reasonable thing to decide that
it is not worth my time to learn how to work with/program on/optimize
on a particular architecture because it is not likely to be
commercially successful.  It is easy enough to be wrong about the
commercial success aspects, but it is an unavoidable question.

>It continues to astounds me how, particularly in the academic computer
>science community, some dodo will stand in front of an audience, show
>that there exists a class of problems which don't run well on parallel
>architectures, and conclude that therefore parallel architectures are
>bad.
>What *frightens* me is that N people will sit in the audience and nod
>their heads in agreement and go out to spread this gospel (as we've
>seen on this list) instead of riding the dodo out on the first rail.

I believe that it is far more common for people to conclude that
parallel architectures are not an effective approach for the class of
problems being discussed.  Since many of us out here are _users_,
rather than designers, it is hardly surprising that we would downplay
the potential usefulness of architectures that are believed to be
unhelpful in our chosen work.  This is quite a reasonable response --
it reminds me of democracy and enlightened self-interest and all of
that stuff.... :-)

>LOOK, there are folks out there using all these architectures and
>winning big. Consider that before you attempt to prove on paper again
>that it's impossible for a honeybee to fly.

It is a good point to note here that lots of the people who think that
parallel architectures are not useful in their field are wrong.  

>But it's so much easier to throw peanut shells from the bleachers.

And so much more fun! ;-)
--
John D. McCalpin - mccalpin@masig1.ocean.fsu.edu
		   mccalpin@scri1.scri.fsu.edu
		   mccalpin@delocn.udel.edu