Re: [boost] Re: Numerical Integration - continued

From: Martin Weiser (weiser_at_[hidden])
Date: 2002-10-23 11:17:57

On Mittwoch, 23. Oktober 2002 17:37, Miroslav Silovic wrote:

> The issue is that this thing makes sense for solving ODEs, but is
> pretty useless for the numeric integration. While numeric integration
> is a special case of solving ODEs, the actual algorithms are completely
> different... the main difference being that ODEs tend to be solved
> stepwise, while functions get integrated in order that isn't even
> necessarily sequential (Romberg algorithm is the primary culprit here).

Sure.  I think that one-dimensional numerical definite integration is a 
sufficiently simple concept and the algorithms are sufficiently mature to 
warrant some easy to use standard interface to be included in boost.

This is completely different for multi-dimensional integration (maybe 
except you restrict it to Cartesian regions in, say, no more than four 
dimensions) and ODE solvers. 

The consequence is, that the design of an "integration interface" for 
boost should be restricted to one-dimensional integration.

Yours
Martin

-- 
Dr. Martin Weiser            Zuse Institute Berlin
weiser_at_[hidden]                Scientific Computing
http://www.zib.de/weiser     Numerical Analysis and Modelling