Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83; site fortune.UUCP
Path: utzoo!linus!decvax!harpo!ihnp4!fortune!rpw3
From: rpw3@fortune.UUCP
Newsgroups: net.lang.c
Subject: Re: Beware: Hackers - (nf)
Message-ID: <2239@fortune.UUCP>
Date: Wed, 11-Jan-84 11:04:03 EST
Article-I.D.: fortune.2239
Posted: Wed Jan 11 11:04:03 1984
Date-Received: Thu, 12-Jan-84 04:19:52 EST
Sender: notes@fortune.UUCP
Organization: Fortune Systems, Redwood City, CA
Lines: 42

#R:allegra:-218900:fortune:16200018:000:1338
fortune!rpw3    Jan 11 04:59:00 1984

I regret to say, my "fix" of the divide-a-negative-variable-by-a-
constant-power-of-two-using-a-shift has a bug in it (OOPS!).

As Gary Graunke (Tektronix) rightly points out, the number that should
be added to the variable before shifting (if the variable was negative)
is the power-of-two-minus-one, not just plain 'ol one.  Now you all
knew I knew all along, right? Well, I did.  (Just ask our local
compiler guys.) but (BLUSH!) I didn't proofread my note very well;
I was too busy with the fancy "learn from history" quotes.  (We will
see next time if I learn from history...)

Rob Warnock

UUCP:	{sri-unix,amd70,hpda,harpo,ihnp4,allegra}!fortune!rpw3
DDD:	(415)595-8444
USPS:	Fortune Systems Corp, 101 Twin Dolphins Drive, Redwood City, CA 94065

+----------------extract from mail--- (also posted to net.lang.c // eventually)
| From allegra!tektronix!tekchips!garyg Tue Jan 10 18:47:51 1984
|
| The incorrect algorithm is...
| 
| 		int x;
| 		unsigned int n;
| 		/* replace x/(2^n), n>=0 by ... */
| 		if(x < 0)
| 		    x += 1;
| 		x = x "arith_right_shift" n ;	/* where the divisor is 2^n */
| 
| The correct algorithm is...
| 
| 		int x;
| 		unsigned int n;
| 		/* replace x/(2^n), n>=0, 2^n-1 <= maxint by ... */
| 		if(x < 0)
| 		    x += 2^n-1;
| 		x = x "arith_right_shift" n ;	/* where the divisor is 2^n */
+---------------