GNU Scientific Library in D

[Don Clugston]

TomD wrote:
> Don Clugston Wrote:
> [...]
>> BTW, it seems to me that the GSL, at least originally, was just a clone of 
>> Numerical Recipes, and the code I've seen reproduces the NR bugs. There is scope 
>> for a lot of improvement, I think.
> GSL was always written in C, exploiting higher level features of a language, whereas
> the Recipes were originally written in FORTRAN, and even the C transcript (I am avoiding
> the word "port" here on purpose) is ugly.

It's even worse than that. Much of the FORTRAN code is actually a port from Algol!

> It is a clone in the sense that it gives tools for
> solving a similar set of problems.

The relationship is closer: it provides the same set of algorithms, organised in 
the same way.

One target actually was to be as complete as the recipes, which it does not do, 
but it had some other additions that the Recipes lack.  Out of curiosity,
> which bugs are in the current version?

Little things like failing to check for overflow. (eg, a-b can overflow if a is 
large and positive, and b is large and negative). Sometimes these errors are not 
present in the original Algol. Extremely poor performance for worst-case 
situations. Mostly pretty obscure stuff.

> 
> Ciao
> Tom