Daniel Barlow dan at
Sun Jan 14 20:28:25 GMT 2007

Jonathan Rockway wrote:
> Simon Cozens wrote:
>> typical, really. Automatic casting to rationals wouldn't be *that* hard.
> Use BigFloat then.  Most people want to do their addition and
> subtraction in registers so that their computations finish some time
> before the heat death of the Universe.  If you want your computer to
> work in decimal instead of binary, then you can wait :)

Most *people*, I suspect, would rather have the right answer later than
the wrong one now - especially as their computer probably isn't doing
very much with the cycles this efficiency saves anyway.  Programmers are
a special case.

How much slower could it be to do the maths with machine words then
check the overflow flag and recompute if it flipped?  More or less slow
than using a bytecompiled language instead of C?  I hear that "Perl"
thing is quite popular these days.


More information about the mailing list