lapack issues



S. Newhouse wrote:
> (%o1)
> (%i2) load(lapack);
>
> Maxima version: 5.13.0
> Maxima build date: 21:33 9/7/2007
> host type: i686-pc-linux-gnu
> lisp-implementation-type: SBCL
> lisp-implementation-version: 1.0.7
>
>
> stdin:16:Incorrect syntax: : is not a prefix operator
>
>   
It actually prints out the maxima build info again?  Any other info 
besides the incorrect  syntax message?  This does seem odd.
> As root, this problem does not happen.  Also, if I put
>   load("/usr/local/share/maxima/5.13.0/share/lapack/lapack.mac");
> in my maxima-init.mac file, then the problem disappears and lapack does 
> load.
>
> Issue #2:  documentation.
>   Is there anyway to get documentation for the maxima version of 
> 'lapack' (other than reading the sources of course)?
>
>   
My apologies.  I've promised Robert that I'd write some documentation, 
but haven't done that yet.  Since people are actually using it now, I'll 
write some very soon.


>  For instance, I looked up the fortran version and one calls
>   dgeev with too many variable to remember.  Maybe it works with fewer 
> variables, I did not try it since I don't use fortran that much anymore.
>
>  I noticed in the archives that
>   'dgeev(M)'
>  produces the eigenvalues of the matrix 'M'.   So, that worked.  But the 
> fortran version docs state that 'dgeevx(M)' gives the eigenvalues and 
> eigenvectors.
>
>   
dgeevx isn't supported.   Do you really need it?

In any case, some simple documentation for dgeev is in the source.  
Basically,

dgeev(matrix [, left-eigen-vec [, right-eigen-vec]])

dgeev all produces the eigenvalues (if possible).  If the optional 
left-eigen-vec is not false, the left eigenvectors are produced.  If 
rigth-eigen-vec is not false, the right eigen vectors are also returned.

I will document these very soon.  There's really no excuse for this.   
Sorry about that.

If you need other functions from lapack, let me know and I can add them 
too.  With documentation, of course. ;-)

Ray