MathGroup Archive 1997

[Date Index] [Thread Index] [Author Index]

Search the Archive

Re: Re: floor problems

  • To: mathgroup at smc.vnet.net
  • Subject: [mg8244] Re: [mg8151] Re: [mg8113] floor problems
  • From: seanross at worldnet.att.net
  • Date: Thu, 21 Aug 1997 21:17:01 -0400
  • Sender: owner-wri-mathgroup at wolfram.com

Lou Talman wrote:
> 
> The example I gave in my earlier post seems to show that multiplication of
> floating point numbers isn't commutative.  In fact, the problem is with
> associativity, as I claimed.  Here is the example I should have given:
> 
> In[2]:=
> ((.7 67 )10)-(.7 (67 10))
> 
> Out[2]=
> \!\(5.68434188608080148`*^-14\)
> 
> --Lou Talman

You are thinking like an abstract mathematician rather than a computer
programmer, scientist or engineer.  For a number that is only has 15
digits of precision, 5.6*10^-14 is "zero".  Multiplication of floating
point numbers is associative--within the precision of the numbers.  You
have to learn to take things like that into account once the mathematics
leaves your pencil and pad and is put on a digital machine like a
computer or calculator.


  • Prev by Date: Re: Re: floor problems
  • Next by Date: Interpolation on irregular grid
  • Previous by thread: Re: Re: floor problems
  • Next by thread: Re: Re: floor problems