RE: [boost] interfaces to numerical functions

From: Paul A. Bristow (boost_at_[hidden])
Date: 2002-11-12 18:21:46

> -----Original Message-----
> From: boost-bounces_at_[hidden]
> [mailto:boost-bounces_at_[hidden]]On Behalf Of Jason D Schmidt
> Sent: Saturday, November 09, 2002 9:06 PM
> To: boost_at_[hidden]
> Subject: [boost] interfaces to numerical functions
>
>
> Hi everyone,
>
> Awhile ago, I posted a few messages about writing some fast Fourier
> transform (FFT) code for Boost.  Now, we have a lot of discussion about
> numerical integration.  If we're going to start writing numerical code
> for Boost (and I hope do a lot of it), we should probably come up with
> some overall guidelines about their interfaces & performance.  Here are
> some of my thoughts on those issues; maybe this could start a useful
> discussion.
>

Jason puts his finger on lots of important problems, some of which have been
aired before,
but without useful conclusion.  Indeed even the simpler question of math
functions (and indeed math constants) seemed to get bogged down without any near
agreement, and some conflict with C99 function naming like double function(),
float functionF(), long double functionL().  I get the impression of a conflict
between C 'diehards' and C++ 'purists'.
With FFT an extra conflict is between speed and C++ niceness.

Some really useful cases using C++ are now looming, for example combining the
uBLAS and interval with integration and similar libraries.

I feel resolving these conflicts is imost mportant - I hope someone can suggest
how.

Paul

Paul A Bristow, Prizet Farmhouse, Kendal, Cumbria, LA8 8AB  UK
[phone]   Mobile [phone]
Mobile mailto:pabristow_at_[hidden]
mailto:pbristow_at_[hidden]