Path: utzoo!utgpu!jarvis.csri.toronto.edu!clyde.concordia.ca!uunet!wuarchive!uwm.edu!rutgers!bpa!cbmvax!mitchell
From: mitchell@cbmvax.commodore.com (Fred Mitchell - PA)
Newsgroups: comp.graphics
Subject: Re: Integral Square Root
Message-ID: <9175@cbmvax.commodore.com>
Date: 31 Dec 89 02:32:13 GMT
References: <9170@cbmvax.commodore.com> <21550@mimsy.umd.edu>
Reply-To: mitchell@cbmvax.commodore.com (Fred Mitchell - PA)
Organization: Commodore, West Chester, PA
Lines: 54

In article <21550@mimsy.umd.edu> chris@mimsy.umd.edu (Chris Torek) writes:
>In article <9170@cbmvax.commodore.com> mitchell@cbmvax.commodore.com
>(Fred Mitchell - PA) writes:
>>Here's one that is short, fast, and sweet.
>
>... and wrong: the line
>
>       if ( (rb = r+(b=1<<i)) * rb <= n)
>
>uses and sets rb in an unknown order.  (Mr. Mitchell's compiler clearly
>does the assignment, then the evaluation of the second `rb', then the
>multiplication, but a compiler is free to evaluate the second `rb', then
>compute r+..., assign that to rb, multiply that by the old value of rb,
>and compare with n.)  The simplest fix is to break up the three embedded
>statements (b=, rb=, if...), which also leads to more easily understood
>code.

NOT WRONG- Let's analyze this line more closely (exploded IF, here we come!)...

	if (
		( rb = r + 
			(b=1<<i) )
				* rb
					<= n)

Now the one stipulation of the C standard is that INNER PARENTHESIS
GETS EVALUATED FIRST!!!! So the first thing that MUST be evaluated
first is:
	(b = 1 << i)

then:
	(rb = r + b)

and finally:
	rb * rb <= n

Don't fault Chris however. It is very terse and a bit hard to understand,
especially for neophyte C programmers. But I wrote this for SPEED and
COMPACTNESS. This arrangement is sure to work on any properly-implemented
C compiler, and if not- then get your money back! Sure you can expand it
if you like- but that defeats the purpose. I have used this simple
routine for years and on several different compilers- even buggy ones-
and have acheived the expected results. This was also originally done
in FORTH and took up twice as many lines of code! (If any are interested
in the FORTH version, beg and plead and I just may dig it up! :-)

>In-Real-Life: Chris Torek, Univ of MD Comp Sci Dept (+1 301 454 7163)
>Domain:	chris@cs.umd.edu	Path:	uunet!mimsy!chris


	-Mitchell
	 mitchell@cbmvax.UUCP
	 "What do you mean I'm wrong? I've been doing this stuff for
 	 twelve years!!!" - Yours Truly.