Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!watmath!clyde!cbatt!ihnp4!inuxc!pur-ee!uiucdcs!uiucdcsm!mccaugh
From: mccaugh@uiucdcsm.UUCP
Newsgroups: comp.lang.c
Subject: Re: math function speed
Message-ID: <4700003@uiucdcsm>
Date: Sat, 21-Feb-87 00:30:00 EST
Article-I.D.: uiucdcsm.4700003
Posted: Sat Feb 21 00:30:00 1987
Date-Received: Mon, 23-Feb-87 02:26:17 EST
References: <2050@ulysses.homer.nj.att.com>
Lines: 16
Nf-ID: #R:ulysses.homer.nj.att.com:2050:uiucdcsm:4700003:000:876
Nf-From: uiucdcsm.cs.uiuc.edu!mccaugh    Feb 20 23:30:00 1987


 A numerical-analyst named McConnell makes chronic contributions to the APL
 Conference Proceedings...in particular, I believe the 1983 edition has an
 article by him definiteively addressing the "square-root" problem. It is
 really quite informative.
 Our own Prof. Gear presents a convincing strategy in his "Computer Organ-
 ization and Programming" (2nd edition) for speeding up square-root compu-
 tation on the IBM-360/370 (he presents the strategy without proof: I have
 a proof in case anyone is interested); not that it is necessarily optimal,
 but it is a "speed-up".
 No, Newton's (actually, Newton-Raphson's) Method is not "optimal": it only
 converges quadratically; for a really nice cubically-converging version,
 look at Franz's "Elements of Numerical Analysis". However, fewer itera-
 tions are bought at the price of more ocmputations per iteration.....