Path: utzoo!utgpu!news-server.csri.toronto.edu!bonnie.concordia.ca!uunet!zaphod.mps.ohio-state.edu!ub!galileo.cc.rochester.edu!rochester!pt.cs.cmu.edu!gandalf.cs.cmu.edu!lindsay
From: lindsay@gandalf.cs.cmu.edu (Donald Lindsay)
Newsgroups: comp.arch
Subject: Analogies and Performance
Message-ID: <13513@pt.cs.cmu.edu>
Date: 19 Jun 91 00:15:16 GMT
References: <2688@m1.cs.man.ac.uk>
Organization: Carnegie Mellon
Lines: 23

In article <2688@m1.cs.man.ac.uk> mshute@cs.man.ac.uk (Malcolm Shute) writes:
>>If long legs are *it*, how come you can't run faster than my dog?
>Does this mean that we should be measuring 'Leg-Complexity', not 'Leg'.

What users measure is "end to end" speed. (That's the track's ends,
not dogends... )

It is interesting to measure factors that contribute to speed, but
there are lots of semantic pitfalls. For example, what does "length"
mean if the leg is never straight?

Measures depend on models, which are abstract. There are usually
problems with abstractions (and with biological analogies). This is
why simulated systems so often perform wonderfully well: they
abstracted out something significant, such as cache misses, or DRAM
refresh, or DMA contention .. or realistic coding practices.

There's a very nice phrase in aerial photography. The photo-analyst
draws deductions from the images: but if she actually goes to the
site, she will learn the "ground truth". It can be an enlightening
experience.
-- 
Don		D.C.Lindsay 	Carnegie Mellon Robotics Institute