Re: [Bug-gsl] Re: [Help-gsl] error in Schur decomposition of a non-symmetric matrix ?

[Patrick Alken]

> The rest of A (below the subdiagonal) is not guaranteed to be 0, as 
> far as I remember.

It just occurred to me, that the rest of A below the subdiagonal is 
probably very close to 0, ie to machine precision due to the way the 
eigenvalue decomposition algorithm works, however at no point are these 
values set identically to 0.0, whereas the subdiagonal elements are set 
to 0.0 in the case of real eigenvalues (this is done to help the 
eigenvector solver detect whether or not they are real or complex 
eigenvalues).

So the bottom of A may contain values like 2.3e-15 or somesuch. The 
function gsl_linalg_hessenberg_set_zero will set these values 
identically to 0.0 if thats what you want.