Re: [Boost-users] Math library on the BOOST licence.

From: John Maddock (john_at_[hidden])
Date: 2006-12-07 11:09:07

Seweryn Habdank-Wojewódzki wrote:
> Hi
>
> John Maddock wrote:
>
>> There is some work in progress here: www.johnmaddock.co.uk/toolkit
>
> Great job. May I download it?

Thanks, there's instructions here: 
http://www.johnmaddock.co.uk/software.html

But we'll also be posting a new release in the Boost Vault (and asking for a 
review in all likelyhood) real soon now...

>> The emphasis is on statistical distributions and special functions,
>> which is rather different from what you're proposing.
>
> My proposition is just an simple example, of the design. If I can say
> something about my design style, I have other more complicated math
> example http://knnl.sf.net and description is
> http://accu.org/var/uploads/journals/overload74.pdf from page 23.
> Unfortunately library is not on the Boost licence. Maybe next version
> I will prepare with Boost licence.
>
> In preparation I have adaptive Runge-Kutta (4,5) method.
>
> I am looking for some tips if there is possible to make generalized
> Runge-Kutta method. In math books there is an algorithm how to
> calculate all coefficients for RK algorithm, but the point is that in
> templates I need to use only integral types, so I need to have all
> coefficients split in to the two parts: numerator and denominator,
> and prepare dedicated algorithm to work with such a fractional
> numbers to calculate RK coefficients in compile time. Otherwise it is
> easy to do algorithm which prepares coefficients in constructor of
> the solver, because double numbers are possible.

There is a library in the Boost Vault that does compile-time math, but I 
haven't had a chance to play with it yet, you might be able to adapt that to 
calculate the coefficients at compile time.  The big problem is likely to be 
avoiding numeric overflow in the fractions, however.

John.