Re: Binding to the GNU Scientific Library

["Dr. Adrian Wrigley" <amtw@linuxchip.demon.co.uk.uk.uk>]

On Fri, 03 Nov 2006 05:09:49 -0800, brian.b.mcguinness wrote:

> 
> Dr. Adrian Wrigley wrote:
> ...
>> And the topic impinges on the question of Ada for scientific programming
>> that was discussed here a few months back.  Lack of efficient library
>> access is a major barrier to use in this application class.
> 
> It seems to me that instead of having a binding to the library it would
> be better to translate  the library into Ada and make it a generic
> package for floating point types.  That would provide more flexibility,
> and would probably be more efficient that having to translate back and
> forth between Ada and C floating point numbers.

Assuming a given level of effort, the benefit would be greater to
bind Ada to an existing (debugged) library than to create one
from scratch.  The time saved could be devoted to improving
the library, benefiting all users.  Or developing your application.
GSL is rather extensive now, and the more it is optimised and extended,
the better.

A well designed library would avoid spending time converting
data formats on the interfaces.  But making this accessible by
generics is a bit tricky because you need to switch beween the
different implementations according to precision, I think.
--
Adrian