Mail Archives: djgpp/1996/12/19/20:28:53
From: | Gary R Sekinger <sekinger+@andrew.cmu.edu>
|
Newsgroups: | comp.os.msdos.djgpp
|
Subject: | Re: floating point/numeric question
|
Date: | Thu, 19 Dec 1996 18:18:34 -0500
|
Organization: | Masters student, Software Engineering, Carnegie Mellon, Pittsburgh, PA
|
Lines: | 28
|
Message-ID: | <UmiQp_600iWm08B340@andrew.cmu.edu>
|
References: | <01ICWAM1K0POAR29Q9 AT mail>
|
NNTP-Posting-Host: | andrew.cmu.edu
|
In-Reply-To: | <01ICWAM1K0POAR29Q9@mail>
|
To: | djgpp AT delorie DOT com
|
DJ-Gateway: | from newsgroup comp.os.msdos.djgpp
|
Excerpts from mail: 11-Dec-96 Re: floating point/numeric .. by H.
Broeker AT physik DOT rwth-a
> yIn article <kmfQlfO00iWn0B9GU0 AT andrew DOT cmu DOT edu> you wrote:
>
> > I plan to write an app and target Unix, dos, and VMS platforms (which is
> > why I picked gcc). The app will produce data based on statistical
> > distributions, so the calculations need to be consistent on all the
> > platforms (but not necessarily accurate :). I've read the floating point
> > portion of the FAQ but still have a question: Does djgpp gcc handle
> > internal calculations and data representations the same as the other
> > distributions of gcc?
>
> Obviously not, because there is no such thing as 'the way gcc handles
> internal calculations', really. That's completely hardware dependant.
> Once you start using floating point numbers, you're subject to design
> decisions of the people who made the floating point unit in that
> machine, like the famous 'Cray style': "who cares if multiplying by an
> exact (!) 1.0 changes the two (!) least significant bits?".
>
> I may have misunderstood, but you seem to be on an impossible mission
> here.
>
> HBB
Thanks for the insight, I think I'm doomed. I kinda thought DJGPP
provided some level of platform independence and provided "source code
compatibility" with its Unix counterparts. BTW, do you know how java
solves this issue?
- Raw text -