Path: utzoo!mnetor!uunet!portal!cup.portal.com!Thomas_E_Zerucha
From: Thomas_E_Zerucha@cup.portal.com
Newsgroups: comp.misc
Subject: Re: Arbitrary precision square root alg?
Message-ID: <3132@cup.portal.com>
Date: 12 Feb 88 01:17:10 GMT
References: <1332@gumby.cs.wisc.edu>
Organization: The Portal System (TM)
Lines: 9
XPortal-User-Id: 1.1001.3273


There is the long-division type method.  You shift two bits (from even
positions) into a working "divisior".  Then the working square root gets a
01 appended (shifted right) into it.  If this value is less than the
working quotient, subtract this remainder from the quotient, and replace
the 01 with a 1, otherwise replace it with a zero.  This sounds complicated
but it is simply the method found in many books converted to binary.  I can
retrieve a more detailed example and post it if you need it.  But you will
need long integers, but you don't need multiplies or divides, just shifts
and subtracts/compares.