
Boost : 
From: Anthony Williams (anthwil_at_[hidden])
Date: 20020823 03:18:41
David Bergman writes:
> Actually,
>
> An implementation is not even allowed to do use that doubleprecision
> "interim".
>
> Since the arithmetic is supposed to be module 2^n, where n is the number
> of bits in the integral type in question.
>
> Remember that "mathematically correct" should actually be
> "mathematically correct in the module 2^n ring".
Only for unsigned types
Anthony
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk