Re: [boost] [SoC] multi precision arithmetic library proposal.

From: Paul A Bristow (pbristow_at_[hidden])
Date: 2006-05-01 07:25:57

| -----Original Message-----
| From: boost-bounces_at_[hidden] 
| [mailto:boost-bounces_at_[hidden]] On Behalf Of John Maddock
| Sent: 01 May 2006 10:42
| To: boost_at_[hidden]
| Subject: Re: [boost] [SoC] multi precision arithmetic library 
| proposal.
| 
| > Yes, I think there would be a positive response to an SOC 
| proposal in
| > this area.      You are correct, there is no multiprecision
| > arithmetic support in Boost now. However, be aware that 
| there is some
| > prior work already in the boost sandbox.  You might want to have a
| > look at these links:
| > Finally, I'll mention that another possible project in this area
| > would be to build a Boost implementation for the Decimal Arithmetic
| > proposal.  This is a proposal that IBM is making before the standard
| > committee.  That latest draft is at:
| >
| > http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n1977.html
| >
| > It would be fabulous to have a Boost implementation of this proposal
| > available.
| 
| Very fabuluous indeed, but have you read the spec?
| It make me quiver just to think about it!

Since 'the student has all summer', I think that decimal is just a little
ambitious!

I think just getting infinite precision integers (and perhaps rationals too)
in Boost would be a good achieveable target.

I find it most diappointing that we have had several implementations in the
sandbox but none have survived a review.  

I am uncertain what higher precision floating-point would be most useful in
Boost.

An exact (infinite precision) real like

http://keithbriggs.info/xrc.html

or a defined precision (typically several hundred decimal digits) like NTL
RR,

or just a higher (128-bit) precision like NTL quad_float, (potentially of
interest to MSVC users who are limited to 64-bit doubles == long doubles
without even an 80-bit option).  Some 'not-so-nice' features of quad_float
have popped-up recently - funny epsilon, thread-unsafe...

(NB Also I believe NTL is GPL not Boost licence).

It may be that these tools are mainly used once only to calculate
coefficients, as John Maddock is doing for 'proper' math functions?  And so
there is less need for these in Boost?

Potential users views?

Paul

-- 
Paul A Bristow
Prizet Farmhouse, Kendal, Cumbria UK LA8 8AB
Phone and SMS text [phone], Mobile and SMS text [phone]
mailto: pbristow_at_[hidden]  http://www.hetp.u-net.com/index.html
http://www.hetp.u-net.com/Paul%20A%20Bristow%20info.html