Path: utzoo!utgpu!water!watmath!uunet!labrea!agate!ucbvax!ihnp4!ihlpg!bgb
From: bgb@ihlpg.ATT.COM (Beuning)
Newsgroups: unix-pc.general
Subject: 3b1 ulrem() bug
Keywords: 3b1 "unsigned long remainder" ulrem bug
Message-ID: <5552@ihlpg.ATT.COM>
Date: 20 Jul 88 00:18:27 GMT
Distribution: unix-pc
Organization: AT&T Bell Laboratories - Naperville, Illinois
Lines: 46

3b1 owners,

	I believe I have found a bug in unsigned long
remainder arithmetic.  If you compile and execute this
program on your 3b1, the value of 'c' is incorrect.


----------------------------------------------------------------
main()
{
	unsigned int	a = 5 * 536870912;
	unsigned int	b = 1073741823;
	unsigned int	c, d;

	c = a % b;
	/* slow but sure method */
	d = a;
	while( d >= b )
		d -= b;

	printf( "a = %u  b = %u  c = %u  d = %u\n", a, b, c, d );

	exit( 0 );
}
----------------------------------------------------------------


The output on my 3b1 is (commas added for readability)

a = 2,684,354,560  b = 1,073,741,823  c = 1,610,612,737  d = 536,870,914

The value of 'd' is correct, the value of 'c' is incorrect.


An "unsigned int" should be able to contain all values
from 0 to 4,294,967,295 (2 ^ 32 - 1).

The line that reads "c = a % b" generates assembly code
that calls the ulrem() function to compute the remainder.
It looks like there is a bug in ulrem().  I have only seen
the problem with very large numbers.



		Brian Beuning
		..!ihnp4ihlpn!bgb