Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!cs.utexas.edu!samsung!rex!ames!dftsrv!mimsy!chris
From: chris@mimsy.umd.edu (Chris Torek)
Newsgroups: comp.graphics
Subject: Re: Integral Square Root
Message-ID: <21561@mimsy.umd.edu>
Date: 31 Dec 89 20:11:08 GMT
References: <9170@cbmvax.commodore.com> <21550@mimsy.umd.edu> <5964@alvin.mcnc.org>
Organization: U of Maryland, Dept. of Computer Science, Coll. Pk., MD 20742
Lines: 11

In article <5964@alvin.mcnc.org> spl@mcnc.org (Steve Lamont) writes:
>[Fred Mitchell's Isqrt is] (apparently) *slower* [than the 4.3BSD VAX sqrt].
>This was including the overhead of converting the argument to a double for
>sqrt() and then converting it back to an int for assignment.

4.3BSD's VAX sqrt uses some fancy VAX code (due to Kahan and his grad
students).  It is MUCH better than the old `portable' double-precision
sqrt, being not only many times faster but also more accurate.
-- 
In-Real-Life: Chris Torek, Univ of MD Comp Sci Dept (+1 301 454 7163)
Domain:	chris@cs.umd.edu	Path:	uunet!mimsy!chris