Path: utzoo!utgpu!jarvis.csri.toronto.edu!rutgers!apple!oliveb!sun!chiba!khb
From: khb%chiba@Sun.COM (chiba)
Newsgroups: comp.arch
Subject: Re: MIPS/MFLOPS ratio [long; here we go again; sorry]
Message-ID: <114186@sun.Eng.Sun.COM>
Date: 7 Jul 89 06:09:13 GMT
References: <596@megatek.UUCP> <112807@sun.Eng.Sun.COM> <114015@sun.Eng.Sun.COM> <37530@sgi.SGI.COM>
Sender: news@sun.Eng.Sun.COM
Reply-To: khb@sun.UUCP (chiba)
Organization: Sun Microsystems, Mountain View
Lines: 26

In article <37530@sgi.SGI.COM> bron@bronze.wpd.sgi.com (Bron Campbell Nelson) writes:

>The surprising thing (to me) is how close the + and * numbers are.  What

It shouldn't be surprising. (If there is interest I can key in
complete op counts for some common kalman filter algorithms, as
examples). Of the infamous BLAS, both dot products and saxpy do one
multiply and one add every time through the innermost loop ... shops
which do serious number crunching typically do stuff like householder
transformations, givens rotations, matrix factorizations, etc. where
the algorithms are so typically close to tied between multiplies and
adds that most folks just count one or the other and multiply by two.

>this unfortunately means is that the answer is not very clear.  It involves
>answering questions like "how frequently can an add be overlapped with
>a multiply?", and "how often is an add on the critical path?"  F.P. adds

These can be overlapped a very large fraction of the time. The easiest
way to see this is to examine common algorithms.

Cheers.

Keith H. Bierman      |*My thoughts are my own. Only my work belongs to Sun*
It's Not My Fault     |	Marketing Technical Specialist    ! kbierman@sun.com
I Voted for Bill &    |   Languages and Performance Tools. 
Opus  (* strange as it may seem, I do more engineering now     *)