From help-octave-request at bevo dot che dot wisc dot edu  Sat Jan 24 04:47:58 1998
Subject: Re: MATLAB COMPATIBILITY (sparse matrices)
From: Ole Myren R|hne <o dot m dot rohne at fys dot uio dot no>
To: help-octave at bevo dot che dot wisc dot edu
Cc: Andy Adler <adlera at njc dot org>
Date: Sat, 24 Jan 1998 11:47:56 +0100

Sorry for the empty message:-|


>>>>> "Adler" =3D=3D Andy Adler <adlera at njc dot org> writes:

    Adler> X=3Ddiag_inv(A,B) A is mxt where t+1 is number of
    Adler> non-zero superdiagonals B is mxn X is mxn

LAPACK handles band matrices:

> DPBTRF (l) - compute the Cholesky factorization of a real symmetric
> positive definite band matrix A
>
> DPBTRS (l) - solve a system of linear equations A*X =3D B with a
> symmetric positive definite band matrix A using the Cholesky
> factorization A =3D U**T*U or A =3D L*L**T computed by DPBTRF
>
> (and many more)

These routines should be well tested & optimized. As Octave already
uses LAPACK for linear algebra one might want to use LAPACK's band
matrix routines as well.

But what about matrices that are sparse but not narrow banded?

	Ole Myren R=F8hne