Problems with linsolve: quotient by polynomial of higher degree
Subject: Problems with linsolve: quotient by polynomial of higher degree
From: Stefano Ferri
Date: Sat, 26 May 2012 11:51:37 +0200
I have written a Maxima program which, given some inputs, automatically
generates systems of linear equations. At the beginning I was using the
Gauss' elimination method, but then I switched to linsolve because of its
greater speed and its capability to produce simpler results.
Unfortunately I've noticed that linsolve cannot be used in all situations:
some systems are quite complicated, and sometimes it returns these errors:
- Quotient by a polynomial of higher degree
- Polynomial quotient is not exact
However, searching in the mailing list archives I've seen the first error
not only affects not only linsolve, but also other commands. A rapid grep
in the source tree showed that both errors are generated by functions
located in rat3a.lisp, the first one in function pquotient (or rquotient),
the second in pquotient1.
I don't know how linsolve works (I don't know lisp and the source code is
uncommented), nor how the above functions work, but the first error makes
me think that there is something related to the search for solutions for
polynomials of degree > 5. Therefore my question is:
Q - Is it possible to use linsolve as a general linear equations systems
solvers, or is it suitable only for solving the simplest cases? Or is this
a bug (in pquotient & c)?
I've attached a file, which can be batched, containing the simplest example
I've found for this problem. There are some polynomials up to the 9-th
order in l, are these the cause of the problem? If l is set to a numeric
value the problem disappears (even though the solution is quite
complicated).
Thanks for any explanation.
Stefano
-------------- next part --------------
A non-text attachment was scrubbed...
Name: linsolve_error.mac
Type: application/octet-stream
Size: 13470 bytes
Desc: not available
URL: <http://www.math.utexas.edu/pipermail/maxima/attachments/20120526/b949680f/attachment-0001.obj>