Re: Floating point problems

[Joachim Wesner <joachim wesner frankfurt netsurf de>]

Richard Henderson wrote:
>
> The question, then, is what to do with the other problematic
> return values, Inf and NaN?  Should Inf be replaced by the
> largest finite value?
> 

IMHO, clearly not. Instead of Denormals, for NaN and Infs, there is
no meaningful replacement value, any program where those appear should
be considered in error and it should trap - as it does already now -
when those occur.

 Joachim