Newsgroups: comp.arch
Path: utzoo!utgpu!watserv1!maytag!watdragon!gcwilliams
From: gcwilliams@watdragon.waterloo.edu (Graeme Williams)
Subject: Re: EFLOP architectures: when and for how much?
Message-ID: <1990Oct29.205812.1641@watdragon.waterloo.edu>
Keywords: pipelining
Sender: gcwilliams@watdragon.waterloo.edu
Organization: University of Waterloo
References: <1dbs02b=035v01@JUTS.ccc.amdahl.com>
Date: Mon, 29 Oct 90 20:58:12 GMT
Lines: 33

In article <1dbs02b=035v01@JUTS.ccc.amdahl.com> rbw00@JUTS.ccc.amdahl.com (  213  Richard Wilmot) writes:
>gcwilliams@watdragon.waterloo.edu (Graeme Williams) said in article
><1990Oct26.191032.9099@watdragon.waterloo.edu> :
>
>> An instruction cannot be executed in a time shorter than the time
>> it takes for a light beam to traverse the processing device.
>
>While it is true that an instruction cannot be EXECUTED in less time
>than light can traverse the computing device, nFLOPS are measures of
>THROUGHPUT.

I don't follow this at all - for a *given* processor is not THROUGHPUT
determined by how fast the processor executes ??

>We can improve throughput by some amount via pipelining of
>the execution so that each section of a computing device performs only
>part of the work for the instruction. After an initial delay to fill the
>pipeline we could produce a result every 1/mth second where this time
>between results < time for light to traverse the whole CPU.

Sure you can do some premliminary work on an instruction using
pipelining techniques - but the bottom line is that the EXECUTION of
an instruction changes the state of the processor (e.g. a register)
and that particular part of the processor involved has a finite
physical size and hence a maximum rate at which it can be changed.
No amount of pipelining can change this fact.

Of course one can do parallel execution (if your algorithm permits)
but then you're no longer using a *SINGLE* processor/execution unit
are you?

Graeme Williams
gcwilliams@watdragon.waterloo.edu