At Wed, 12 Jul 2006 19:23:48 -0500,
Duane Nykamp wrote:
> The function gsl_linalg_solve_symm_tridiag is documented (and named) as
> a solver for a general symmetric tridiagonal system.  It uses a variant
> of Cholesky decomposition, treating the diagonal differently so that the 
> algorithm applies to more than just positive definite matrices.
> 
> However, if that diagonal in this decomposition is ever zero (e.g., the 
> upper-left entry of the original matrix is zero), the algorithm will 
> divide by zero without checking, producing NaNs.

Thanks for the bug report.  I have logged it in the BUGS file.  

For now I have added a note to the documentation that the routine
can fail if the matrix is not positive definite.

-- 
best regards,

Brian Gough

Network Theory Ltd,
Publishing the GSL Manual - http://www.network-theory.co.uk/gsl/manual/


_______________________________________________
Bug-gsl mailing list
[email protected]
http://lists.gnu.org/mailman/listinfo/bug-gsl

Reply via email to