Re: normalized ALF (Assotiated Legendre Function)

[Ben Abbott]

On Feb 12, 2008, at 1:38 PM, Ben Abbott wrote:

> On Tuesday, February 12, 2008, at 10:02AM, "Marco Caliari" <address@hidden 
> > wrote:
> > Hi.
> >
> > The normalized Lagrange functions should (almost) never give NaNs  
> > of Infs.
> > The enclosed script fixes a problem in the previous:
> >
> > octave:1> legendreold(151,-0.9,"norm")(end-1:end)
> > ans =
> >
> >  -3.3248e-53
> >          Inf
> > octave:2> legendre(151,-0.9,"norm")(end-1:end)
> > ans =
> >
> >  -3.3248e-53
> >   9.2660e-55
> >
> > Marco
>
> I've tried to combine the various implementations. I still need to  
> look into adding a warning in then event of an overflow.
>
> We should also add more tests to include the current improvements.
>
> Ben

FYI,

I asked Jason Riedy to run the example below for me with his patch/ 
workaround for xgelsd

        legendre  (80, [-1:0.1:1]);

I have been getting the warning below.

        warning: dgelsd: rank deficient 21x81 matrix, rank = 7

With the patch/workaround Jason did not get the warning.

Ben