Re: Bug report: double floating point arithmetic errors under Linux

From: Warwick Harvey <wh_at_icparc.ic.ac.uk>
Date: Tue 22 Oct 2002 11:05:48 AM GMT
Message-ID: <20021022120546.D7587@tempest.icparc.ic.ac.uk> 
SPARC:
> [eclipse 1]: N is breal(0_1).
> 
> N = -4.94065645841247e-324__4.94065645841247e-324
> Yes (0.00s cpu)

Linux:
> [eclipse 1]: N is breal(0_1).
> 
> N = -2.2250738585072014e-308__2.2250738585072014e-308
> Yes (0.00s cpu)

This one is because Linux defines MINDOUBLE to be the smallest normalised
double, whereas SPARC defines it to be (presumably) the smallest
denormalised double.  And the ieee_up/down code uses MINDOUBLE as the result
if the input is zero.

I'm still looking into the rest of it.

Warwick
Received on Tue Oct 22 12:05:48 2002

This archive was generated by hypermail 2.1.8 : Wed 16 Nov 2005 06:08:18 PM GMT GMT