This is the mail archive of the
gsl-discuss@sources.redhat.com
mailing list for the GSL project.
Re: gsl_linalg_solve_symm_cyc_tridiag problem
- From: Brian Gough <bjg at network-theory dot co dot uk>
- To: David Necas (Yeti) <yeti at physics dot muni dot cz>
- Cc: gsl-discuss at sources dot redhat dot com
- Date: Wed, 17 Apr 2002 20:06:31 +0100 (BST)
- Subject: Re: gsl_linalg_solve_symm_cyc_tridiag problem
- References: <20020406141543.B7885@physics.muni.cz><15536.44975.747008.972898@debian><20020414180221.A10319@physics.muni.cz>
David Necas (Yeti) writes:
> And, to be at least a little constructive: I've implemented
> nonsymmetric tridiagonal system solvers (mainly because I don't
> understand gsl_linalg_solve_symm_cyc_tridiag() and thus can't fix
> it ;-) -- the non-cyclic is almost identical to
> gsl_linalg_solve_symm_tridiag(), only doesn't require symmetrical
> matrix, the cyclic solver uses Sherman-Morrison formula to
> compensate for the offending corner elements after solving the
> system w/o them. A patch is attached, if you'd like to include it.
> Note these routines have problems with too small N too, but it's
> easy to fix them once one decides what to do in case of too small
> N.
>
Thanks -- I've added those in now. Some tridiag test cases would be
useful if you have the chance to work on it.
Brian