Discriminant <0 error

[markhl]

On Aug 24 near 1 PM PDT I got this error 5 times on my home PC, Windows ME, Dell 833 MHz:

ERROR: [001.013] {randwalk.c, line 3878} Discriminant <0 (-0.000000) while calculating CB

This error did not reappear.

Howard's previous posts suggest that this is due to overheated or failing hardware. That may be right in this case; it was a warm Southern California day. I have never run FPU tests on this box, but its RAM passed an overnight memory test about 2 months ago.

Mark

[bwkaz]

Well, the funny thing about IEEE 754 (is it 754? I believe so...) floating-point numbers is that negative zero does actually exist (for those that know what it means, they're basically sign-magnitude). I'm pretty sure it's functionally equivalent to zero anyway.

Or, you might be seeing roundoff errors, and the determinant might be like -0.00000012 or something.

My first guess is that it's harmless -- but I'd say wait for a reply from Howard before dismissing the errors.

[Brian the Fist]

Yes you are right. It is probably just a fluke and can be ignored. If it continues to be a problem, it could indeed be rounding error from the FPU.

The program continues on despite the error so consider it just a warning.