Re: floor problems
- To: mathgroup at smc.vnet.net
- Subject: [mg8288] Re: [mg8140] [mg8113] floor problems
- From: "P.J. Hinton" <paulh>
- Date: Sun, 24 Aug 1997 04:46:28 -0400
- Organization: Wolfram Research, Inc.
- Sender: owner-wri-mathgroup at wolfram.com
Sean Ross wrote: > Will Self wrote: > > > > Hi Sean, I just resubscribed. I think that any reasonable human being > > should reasonably expect that Mathematica should give the correct > > answers for *both* of those Floor expressions, as indeed my old > > version 2.0 does. It is all very well to exhort people to be careful > > and be extra-aware, but if one has to go around walking on eggs, then > > what becomes of the usefulness of Mathematica? I say, let the blame > > rest where it should, on WRI. > > I always "walk on eggshells" and never trust a numerical coprocessor or > a compiler further than I can spit. We just had a post where someone > with a Unix based system does not have the error that your powerPC chip > and my Pentium chip do. I don't think this kind of error has anything > to do with mathematica, but with the way the numeric coproccessor > handles floating point and integer conversions. I think the wolfram > people were just talented in exposing the error in the numeric > coprocessor. Anyone who would ever totally and blindly accept the results that come from a numerical algorithm deserve what he or she gets. For an interesting discussion on this, read the book _Real Computing Made Real_ by Forman Acton (except that section in the Exhortation titled "Should I use Mathematica to solve my problem?" :-) ). Very enlightening stuff, indeed. -- P.J. Hinton Mathematica Programming Group paulh at wolfram.com Wolfram Research, Inc. http://www.wolfram.com/~paulh/ Disclaimer: Opinions expressed herein are those of the author alone.