From: Michael Stevens (mail_at_[hidden])
Date: 2005-08-26 10:42:29
On Friday 26 August 2005 16:20, Peter Schmitteckert wrote:
> Salut Michael,
> On Fri, 26 Aug 2005, Michael Stevens wrote:
> > Alternatively it may be possible to solve the problem by implementing
> > uBLAS so the zero elements are not constructed in any algorithms. Using
> > the operation Z = 0* A where 0* multiples elements of A by the scalar 0
> > we can generate Z such that its assignment and additively conformant with
> > A.
> You can't use this to initialize a vector to zero,
> x = 0 * y
> is undefined for an unitialized y, and if it's an array element
> it could contain a NaN and 0 * Nan = Nan and not zero.
Thanks for pointing out the NaN problem. So much for that plan!