From bug-octave-request  Fri Jul 30 10:57:37 1993
Subject: expm
From: A. Scottedward Hodel <Dr dot A dot Scottedward dot Hodel at eng dot auburn dot edu>
To: bug-octave
Date: Fri, 30 Jul 93 10:57:33 CDT


John,

I just looked at your code for the matrix exponential.

Please pardon my french, but *BLEAH*!

I am going to rewrite this code to use Ward's algorithm (1977), 
pretty much accepted as the best general purpose workhorse.  In cases
where the denominator matrix is ill conditioned, it may be worthwhile
to do a check and switch to preconditioned Taylor series with scaling/squaring
instead, but I'm going to leave that alone.

I won't get this done before I leave twon for a week (playing music in
a wedding), but I will certainly get it done ASAP, along with the
generalized eigenvalue code I wrote awhile ago.

[I got run over the coals on the matrix exponential a couple of years ago
while writing sampled-data system simulation code at MSFC, so this is
one of my "hot-buttons", I guess.]

Hope things are going well bringing up your machines.

Scotte