Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!husc6!hao!gatech!hubcap!delp
From: delp@udel.EDU (Gary Delp)
Newsgroups: comp.hypercube
Subject: Re: Amdahl's Law
Message-ID: <491@hubcap.UUCP>
Date: Tue, 22-Sep-87 08:25:47 EDT
Article-I.D.: hubcap.491
Posted: Tue Sep 22 08:25:47 1987
Date-Received: Thu, 24-Sep-87 04:42:34 EDT
Sender: fpst@hubcap.UUCP
Lines: 29
Keywords: parallel, amdahl, laws
Approved: hypercube@hubcap.clemson.edu
Summary: Indeed it is a law, anyone with more references?

in article <11370002@hpsmtc1.HP.COM>, swh@hpsmtc1.HP.COM (Steve Harrold) says:
> If there is some dispute as to how accurate Amdahl's "Law" is, shouldn't
> it be more properly labelled "Amdahl's Theorem"??

Indeed, what people have said is that the world of parallel processing
isn't as bad as Amdahl's Law might first make it seem, not that the
"law" is not true.  They have said that the serial portion of code can
be minimized and separate tasks may be run in parallel.  But, no one
has given an example of where the *law* (like the *law* of gravity, an
observation which fits all the observed data and predicts nature (also
like the law of gravity, arbitrarily complex depending on what factors
must be considered, but simple in the generally used cases.))  has
been shown to be wrong.

The law, simply stated, says that if you need to do something, and
that thing has some elements which depend on their happening in a
certain order, then it doesn't matter how many widgets there are to do
that thing, the speed with which it can be done is limited by the
serial execution speed of the things that must be done in series.

BTW, the reference that eugene miya graciously provided (thanks
eugene) only vaguely refered to this law.  Is it possible that there
are more recent references?  Has anyone heard him talk and refer to
this law?  Any information would be helpful.
-- 
Gary Delp	     ARPA:   delp@udel.edu CSNET:  delp%udel.edu@relay.cs.net
123 Hall, EE U of D		  ...!ihnp4!berkeley	  -\
Newark, DE 19716     UUCP:	  ...!allegra!berkeley	  -->!delp@udel.edu
(302)-451-6653 or 2405		  ...!harvard		  -/