Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!sun-barr!lll-winken!elroy.jpl.nasa.gov!usc!zaphod.mps.ohio-state.edu!rpi!sarah!bingnews!kym
From: kym@bingvaxu.cc.binghamton.edu (R. Kym Horsell)
Newsgroups: comp.arch
Subject: Divide in 1 cycle (was Re: standard extensions)
Message-ID: <1991Mar1.033353.28445@bingvaxu.cc.binghamton.edu>
Date: 1 Mar 91 03:33:53 GMT
References: <1991Feb25.135057.23667@linus.mitre.org> <1991Feb27.021435.11296@bingvaxu.cc.binghamton.edu> <6968@mentor.cc.purdue.edu>
Organization: State University of New York at Binghamton
Lines: 13

In article <6968@mentor.cc.purdue.edu> hrubin@pop.stat.purdue.edu (Herman Rubin) writes:
[there pipes that divide in 1 cycle]
>Of the machines I know, this is not the case.  The CYBER 205, which 
>pipelines just about everything, does not pipeline divide, square root
>(the same unit), or convert between decimal and binary, which I believe
>also uses that unit.  The CRAY does not provide a division at all, presumably
>because it could not be pipelined.  It does have a pipelined approximation to
>reciprocal, and a pipelined correction factor.

I know the Cray-2 has problems with chaining (:-), but on the X-MP
wasn't it possible to chain recip & multiply? 

-kym