Re: [boost] Fixed-Point Decimal Type

From: Beman Dawes (bdawes_at_[hidden])
Date: 2001-05-09 14:48:04

At 02:45 PM 5/9/2001, John Max Skaller wrote:

 >> - All operations are exact if there is no overflow and if no
 >>   rounding is required.  Overflow is undefined behavior.
 >>   Rounding is done the way accountants do it:  round to nearest
 >>   with 1/2-LSB always rounded "up."  (I think "up" means "away
 >>   from zero."  I've asked an accountant friend via e-mail but
 >>   haven't yet gotten an answer.)
 >
 >	This is problematic. For some operations, it is
 >wrong to always round 1/2 LSD up: when doing statistical
 >operations, rounding up or down at random is correct.
 >Sometimes, truncation is correct (rather than rounding).
 >
 >	This suggests that you pick one mode (as you have)
 >for the operator?? forms, but supply the other operations
 >as named functions, for example
 >
 >	mul_with_truncate(a,b)
 >	mul_with_rhlsd_up(a,b)
 >	mul_with_rhlsd_ran(a,b)
 >	mul_with_noround(a,b)

Maybe.  But maybe just stick with one form of rounding, unless there really 
is some overwhelming need for additional forms.

 > ...
 >
 >	It is also OK to leave this stuff out: a basic
 >library is better than none, and much better than
 >getting it completely wrong :-)

I second John's comment:-)

--Beman