NetNews Offline 2/NetNews Offline Volume 2.iso/news/comp/lang/c++-part1/881

Internet Message Format | 1996-08-06 | 1.6 KB

Path: ix.netcom.com!netnews
From: [e-mail] (Mike Rubenstein)
Newsgroups: comp.lang.c++
Subject: Re: Error?
Date: Mon, 08 Jan 1996 02:03:04 GMT
Organization: Netcom
Message-ID: <[e-mail]>
References: <4cpaov$[e-mail]>
NNTP-Posting-Host: ix-dc14-13.ix.netcom.com
X-NETCOM-Date: Sun Jan 07  6:02:44 PM PST 1996
X-Newsreader: Forte Agent .99c/16.141

[e-mail] (Lars Kaderali) wrote:

|>Hey folks! Need some help - I tried the following program in c++ on
an
|>ibm pc:
|>
|>=======================================
|>
|>//test
|>#include <iostream.h>
|>float a,b;
|>main()
|>{
|>  a=-1.0; b=0.1;
|>  while (a<1.1)
|>  {
|>    cout<<a<<endl;
|>    a+=b;
|>  }
|>}
|>
|>=======================================
|>
|>which returns:
|>
|>-1
|>-0.9
|>-0.8
|>-0.7
|>-0.6
|>-0.5
|>-0.4
|>-0.3
|>-0.2
|>-0.1
|>(so far as expected, but then:)
|>7.450581e-08     < ==== ??????????????
|>0.1
|>0.2
|>0.3
|>0.4
|>0.5
|>0.6
|>0.7
|>0.8
|>0.9
|>1
|>
|>How come this 7.450581e-08 happens instead of 0.0 ?????

Welcome to the world of binary floating point.  On most machines
floating point numbers are stored in binary.  In binary 0.1 cannot be
expressed exactly with a finite number of digits.  The computer does
the best it can, but actually uses a number that's just close.

This can cause other anomolies.  In some cases your loop may go one
step further or less than you expect.

Using double will make the numbers more accurate, but still not exact
and the same kind of think will happen.  Here, the number printed
where you expect 0 will be smaller, but probably still not 0.

Michael M Rubenstein