At Wed, 9 Sep 2009 16:18:54 +0100, Jonathan Underwood wrote: > It'd be nice if it was possible to implement an approach such as > CESTAC/CADNA in GSL ..... however, I don't see how to do it in C, as > it seems to require overloading of the basic operators (+/-* etc).
Yes that would be the ideal way to do it, with the error estimate determined automatically, but as you suggest C is not capable of handling that--it would have to be done by code generation or a parser that preprocesses the source (which is how this sort of thing used to be done in FORTRAN 77 I believe). I found the SIAM book "Accuracy and reliability in scientific computing" a useful reference on this subject, although it is a few years old now so it may not cover the latest developments. -- Brian Gough
