Re: Non-deterministic numerical inaccuracies in
- To: mathgroup at smc.vnet.net
- Subject: [mg94952] Re: [mg94908] Non-deterministic numerical inaccuracies in
- From: Bob Hanlon <hanlonr at cox.net>
- Date: Wed, 31 Dec 2008 06:10:40 -0500 (EST)
- Reply-to: hanlonr at cox.net
$Version
7.0 for Mac OS X x86 (64-bit) (November 11, 2008)
No problem apparent.
Bob Hanlon
On Tue, Dec 30, 2008 at 12:26 PM , Sjoerd C. de Vries wrote:
> Hi all,
>
> Last week Budgemano posted an issue which involved numerical
> inaccuracies in Mathematica 7. Due to the xmas season it hasn't
> received much attention, or so it seems.
>
> Meanwhile, I've been able to drill the problem down to the following
> code:
>
> myMod := Module[
> {p},
> Dot[{0.5018839897852984567457648, 0.544293499215831476457645674576,
> 0.67220312964800834576456745674567}, {-0.5359693986690249`, \
> -0.41257352636613975`, -0.736559494563862`}]*{-0.5359693986690249`, \
> -0.41257352636613975`, -0.736559494563862`} - {0.5579131132707782`,
> 0.2715072333003815`, 0.7842300557273234`}]
>
> For[ii = 1, ii <= 250, ii++,
> If[myMod =!= myMod,
> Print["Iteration " <> ToString[ii] <> " fails comparison "]];]
>
> As you can see the module myMod is called twice every iteration. Its
> calculations only involve constants, so the results should always be
> the same. However, on my system (Mathematica 7.WinXP) the results
> differ from one
> call to the other at seemingly random times.
>
> This phenomenon crucially depends on the presence of the local
> variable p (that isn't used!). With this variable it's there, without
> it isn't. In M6 everything is OK, so this is a bug specific to
> Mathematica 7.
>
> Could you check on the various platforms whether you can reproduce
> this?
>
> Cheers -- Sjoerd