On Wednesday 17 August 2005 22:56, Jim McGowen wrote:
> I wrote a test routine that multiplies and casts every number from 0.0001
> to 0.9999 and outputs to a text file via a conduit for analysis.  I haven't
> sifted through the whole set yet but so far I found these to be "trouble
> numbers":
> 0.141 - 0.143
> 0.172 - 0.174
> 0.204
> 0.284 - 0.286
> 0.344 - 0.349
> 0.407 - 0.409
> 0.563 - 0.573
>
> I was thinking I might find a pattern and maybe special case these numbers
> but I'm not smart enough to see a pattern here.  [...]

There is a text "What Every Computer Scientist Should Know About 
Floating-Point Arithmetic" by David Goldberg somewhere in the web, but 
unfortunately I have lost the URL. It explains (among tons of other stuff) 
why certain numbers can't be represented in the expected precision: it has 
all to do with the binary system...

Bodo


-- 
For information on using the PalmSource Developer Forums, or to unsubscribe, 
please see http://www.palmos.com/dev/support/forums/

Reply via email to